{"source":"https://recipes.branolabs.com","license":"CC BY 4.0 for AgentRecipes summaries and prompt templates; linked source write-ups belong to their authors","updated":"2026-10-15","schema_version":1,"schema":{"slug":"URL slug (unique)","title":"How-to title","category":"category slug","platform":"main agent/automation platform used in the source","tools":"tools named in the source","problem":"problem statement (our words)","steps":"step-by-step recipe summarised from the source (our words)","prompt_template":"original prompt TEMPLATE written by AgentRecipes, not taken from the source; fill in [BRACKETS] and {{variables}}","watch_outs":"practical caveats reported in the source","reported_result":"what the source author reports; claims are theirs and unverified by us","source":"the real-world original (url, site, published date or null)","via":"where we found it (discovery credit)","added":"date added","last_checked":"date the source link was last checked"},"categories":[{"slug":"customer-support","name":"Customer support","description":"Recipes for triaging support email, answering customers on WhatsApp and chat, and escalating the hard cases to humans."},{"slug":"sales-leads","name":"Sales & lead generation","description":"Recipes for qualifying leads, researching prospects and meeting guests, and drafting outreach and follow-ups with a human in the loop."},{"slug":"finance-invoicing","name":"Finance & invoicing","description":"Recipes for extracting invoice data, chasing overdue payments, reconciling bank statements and finding forgotten subscriptions."},{"slug":"inbox-calendar","name":"Inbox & calendar","description":"Recipes for triaging email, unsubscribing in bulk, turning bookings into calendar events and getting a morning briefing."},{"slug":"research-monitoring","name":"Research & monitoring","description":"Recipes for summarising newsletters, papers, videos and Reddit threads into Notion or Docs, and for weekly competitor monitoring."},{"slug":"content-social","name":"Content & social media","description":"Recipes for turning blog posts, newsletters and spreadsheet ideas into social posts, with approval before publishing."},{"slug":"team-ops","name":"Team operations","description":"Recipes for async standups, project digests, sprint changelogs and sorting large piles of documents."},{"slug":"engineering-security","name":"Engineering & security","description":"Recipes for AI pull-request reviews, CVE monitoring and Kubernetes alert diagnosis."}],"recipes":[{"slug":"triage-support-emails-draft-replies-n8n","title":"How to triage support emails and draft replies with n8n","category":"customer-support","platform":"n8n","tools":["n8n","Gmail","OpenAI-compatible LLM (gpt-4o-mini in the shared workflow)","Google Sheets","Slack"],"problem":"A shared support inbox mixes urgent problems with routine questions. Reading, sorting and drafting an answer for every message eats hours, but letting a bot reply on its own is risky.","steps":["Add a Gmail trigger in n8n that polls the support inbox for new messages.","Pass each email's sender, subject and body to an LLM node and ask for structured JSON: category, urgency, a one-sentence summary and a short draft reply.","Parse the JSON and append one row per email to a Google Sheet so you have a log.","Post the summary, urgency and draft to a Slack channel for the support team.","Keep sending manual: a person checks the draft, edits it and replies from Gmail."],"prompt_template":"You are a support triage assistant for [COMPANY]. Read the email below and reply with JSON only, using these keys:\n- category: one of [billing, bug, how-to, account, other]\n- urgency: low | medium | high\n- summary: one sentence\n- draft_reply: 3-5 friendly sentences in [LANGUAGE]; do not promise refunds, discounts or deadlines\nIf the message is spam or not a support request, use category \"other\" and an empty draft_reply.\n\nFrom: {{from}}\nSubject: {{subject}}\nBody: {{body}}","watch_outs":["The shared Code node only processed the first email of each poll; loop over every item instead.","Read the sender's text field rather than the raw sender object.","Add JSON-parsing error handling and a spam filter before using it in production."],"reported_result":"The builder says the five-node workflow took about 20 minutes to wire up and that the drafts were good. No volume or time-saved figures were shared, and community reviewers found bugs in the posted version.","source":{"url":"https://community.n8n.io/t/ai-support-inbox-triage-classify-prioritize-draft-the-reply-human-still-sends/311971","site":"n8n Community forum","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/triage-support-emails-and-draft-replies-for-human-review","title":"Triage support emails and draft replies for human review"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/triage-support-emails-draft-replies-n8n"},{"slug":"auto-resolve-support-tickets-n8n","title":"How to auto-resolve support tickets and escalate the rest with n8n","category":"customer-support","platform":"n8n","tools":["n8n","OpenAI (GPT-4o-mini)","SMTP email","Google Sheets","Slack","Optional: knowledge-base and order-history APIs"],"problem":"Many tickets can be answered from your docs or the customer's order data, while billing and security issues need a person quickly. Doing that split by hand slows everyone down.","steps":["Receive tickets on an n8n webhook from your contact form, chat widget or API.","Run a knowledge-base search and an order-history lookup in parallel.","Ask the model to classify category, priority and sentiment, and to draft a reply grounded in what was retrieved.","If the ticket is resolvable, send the reply as an HTML email, log it in Google Sheets and post a note to Slack.","Otherwise route it to a category-specific Slack channel with the draft, a confidence score and notes. Always escalate billing and refund tickets, and send account-security issues to an urgent channel."],"prompt_template":"You are a first-line support agent for [COMPANY]. Use ONLY the knowledge-base excerpts and order data below. Decide whether you can fully resolve the ticket.\nReturn JSON: {\"category\": \"...\", \"priority\": \"low|normal|urgent\", \"sentiment\": \"...\", \"can_resolve\": true|false, \"confidence\": 0-1, \"draft_reply\": \"...\", \"notes_for_human\": \"...\"}\nRules: billing, refund and account-security tickets always get can_resolve = false. Never invent policies that are not in the excerpts.\n\nTicket: {{ticket}}\nKnowledge base: {{kb_results}}\nOrders: {{orders}}","watch_outs":["Hard-code escalation for money and security topics instead of leaving it to the model.","You need OpenAI, SMTP, Google Sheets and Slack credentials before anything runs."],"reported_result":"The author reports very low running costs (about $0.90 a month at 50 tickets a day on GPT-4o-mini, by their estimate). The post gives no resolution-rate or accuracy figures.","source":{"url":"https://community.n8n.io/t/ai-customer-support-agent-auto-resolves-tickets-escalates-to-humans/285840","site":"n8n Community forum","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/auto-resolve-support-tickets-and-escalate-the-rest","title":"Auto-resolve support tickets and escalate the rest"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/auto-resolve-support-tickets-n8n"},{"slug":"whatsapp-support-agent-human-handoff-n8n","title":"How to answer WhatsApp customer questions with an AI agent and hand off to humans (n8n)","category":"customer-support","platform":"n8n","tools":["n8n","WhatsApp Business API","OpenAI","Telegram (for the human team)","Your knowledge base"],"problem":"Customers message on WhatsApp at all hours with the same questions, but a bot that guesses wrong can do more harm than no reply at all.","steps":["Trigger an n8n workflow on every incoming WhatsApp message.","Load the conversation history and the relevant knowledge-base content.","Have OpenAI draft a reply based only on that content, and run a confidence check on it.","If the answer is confident, send it back on WhatsApp.","If not, forward the case to your team's Telegram chat with the full conversation attached so a person can take over."],"prompt_template":"You answer WhatsApp messages for [BUSINESS]. Use only the knowledge base below. Reply in the customer's language in at most 3 short sentences.\nThen, on a new line, write CONFIDENCE: high or CONFIDENCE: low. Use low if the answer is not clearly in the knowledge base, the customer seems upset, or the topic is [refunds / complaints / legal].\n\nConversation so far: {{history}}\nKnowledge base: {{kb}}\nNew message: {{message}}","watch_outs":["The source does not say how confidence is scored, so you have to design and test that check yourself.","Start with a small, accurate knowledge base; the agent can only be as good as its content."],"reported_result":"The repository describes the architecture and its goals (cover routine questions, give round-the-clock coverage, keep humans on unclear cases). It reports no measured results, and the exported workflow was still to be added.","source":{"url":"https://github.com/zezo640/whatsapp-ai-support-agent-n8n","site":"GitHub","published":"2026-07-01"},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/answer-whatsapp-customer-questions-and-escalate-to-humans","title":"Answer WhatsApp customer questions and escalate to humans"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/whatsapp-support-agent-human-handoff-n8n"},{"slug":"triage-repair-requests-airtable-n8n","title":"How to triage service requests into Airtable and auto-reply with n8n","category":"customer-support","platform":"n8n","tools":["n8n","Web form","LLM","Airtable","Email"],"problem":"A local service business (the source example is a car repair shop) gets web-form requests of every kind. Sorting them by job type and urgency and replying quickly is manual work.","steps":["Send each web-form submission to an n8n webhook.","Have the AI classify the request by job type and urgency and write a short summary.","Create an Airtable record with those fields plus a status.","Send the customer an instant reply acknowledging the request.","Flag uncertain cases for human review, and send a polite automatic decline when the request is out of scope."],"prompt_template":"You triage requests for [SHOP NAME], a [TYPE OF BUSINESS] that offers: [LIST OF SERVICES].\nReturn JSON with: job_type (one of the services above or \"other\"), urgency (today | this week | flexible), summary (one sentence), in_scope (true/false), needs_human (true if you are unsure about anything), customer_reply (2-3 sentences; if out of scope, decline politely and suggest [ALTERNATIVE]).\n\nRequest: {{form_data}}","watch_outs":["Write down exactly which jobs are in scope, otherwise the decline rule will misfire.","Check the human-review queue at least daily so flagged customers aren't left waiting."],"reported_result":"The demo shows an Airtable base tracking job status and timestamps for processed requests. No metrics were shared.","source":{"url":"https://x.com/ho_lar_s/status/2109790708970438953","site":"X","published":"2026-10-12"},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/triage-car-repair-requests-and-reply-to-customers","title":"Triage car repair requests and reply to customers"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/triage-repair-requests-airtable-n8n"},{"slug":"qualify-leads-human-approval-n8n","title":"How to score leads and draft outreach emails with human approval in n8n","category":"sales-leads","platform":"n8n","tools":["n8n","LLM","Gmail","Google Sheets"],"problem":"Scoring each new lead and writing a first email by hand is slow, but fully automatic outreach risks sending bad emails to real prospects.","steps":["Receive the lead from a form or webhook.","Validate the fields and check the sheet for duplicates.","Ask the AI to score the lead HOT, WARM or COLD and draft a personalised email.","Write the lead, score, draft and an approval status to Google Sheets.","When a person marks the row approved, send the draft with Gmail and update the status."],"prompt_template":"You qualify inbound leads for [COMPANY], which sells [OFFER] to [IDEAL CUSTOMER].\nScore the lead HOT, WARM or COLD using these criteria: [CRITERIA].\nReturn JSON: {\"score\": \"...\", \"score_reason\": \"one sentence\", \"next_step\": \"...\", \"email_subject\": \"...\", \"email_body\": \"...\"}\nThe email must be under 120 words, mention one specific detail from the lead, and end with one clear question. Do not invent facts about the lead.\n\nLead: {{lead}}","watch_outs":["Keep approval inside Google Sheets at first and only build a dashboard once you know which fields reviewers actually use.","Store a short reason next to every score so reviewers can trust or correct it."],"reported_result":"Shared as a V1 showcase with no results or numbers. A commenter suggested adding a score-reason field so approvers see source, reason, next step and draft in one row.","source":{"url":"https://community.n8n.io/t/i-built-leadflow-ai-v1-an-ai-lead-qualification-workflow-with-human-approval/309312","site":"n8n Community forum","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/qualify-leads-and-draft-emails-pending-human-approval","title":"Qualify leads and draft emails pending human approval"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/qualify-leads-human-approval-n8n"},{"slug":"gmail-leads-to-notion-crm-zapier-agents","title":"How to turn new Gmail emails into researched Notion CRM leads with Zapier Agents","category":"sales-leads","platform":"Zapier Agents","tools":["Zapier Agents","Gmail","Web search","Notion"],"problem":"Potential customers are buried among ordinary email. Copying them into a CRM and researching each sender is tedious, so leads slip through.","steps":["Create a Zapier Agent behaviour triggered by Gmail's New Email event.","In plain-language instructions, tell it to decide whether the email is a lead.","If it is, have it research the sender online (company, website, LinkedIn).","Add an item to your Notion CRM database with mapped properties and the email thread URL in the notes.","Test on sample emails, refine the instructions, then switch it on."],"prompt_template":"For each new email:\n1. Decide whether the sender is a potential customer for [BUSINESS]. Signs of a lead: [asks about pricing, services, availability]. Ignore newsletters, receipts, notifications and existing clients from [DOMAINS].\n2. If it is a lead, look up the sender's company and website.\n3. Create a Notion item with Name, Company, Website, Need (one sentence), Source = \"Email\", Notes = thread URL plus a two-line research summary.\nIf it is not a lead, do nothing. If research finds nothing reliable, write \"unknown\" rather than guessing.","watch_outs":["Each run used about 4 Zapier Agents activities, which is how usage is billed.","A classic Zap is more predictable; an agent needs careful prompt tuning."],"reported_result":"Once switched on it runs on every new email like a normal Zap. The author gives no accuracy or time-saved figures and notes the use case may not suit every business.","source":{"url":"https://www.simonesmerilli.com/business/zapier-agents","site":"Simone Smerilli's blog","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/turn-new-gmail-emails-into-notion-crm-leads","title":"Turn new Gmail emails into Notion CRM leads"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/gmail-leads-to-notion-crm-zapier-agents"},{"slug":"research-meeting-guests-slack-briefing-zapier-agents","title":"How to research meeting guests and post a daily Slack briefing with Zapier Agents","category":"sales-leads","platform":"Zapier Agents","tools":["Zapier Agents","Google Calendar","Slack","Zapier's built-in web search"],"problem":"Before calls with external people someone has to look up who they are and what their company does, usually in a rush right before the meeting.","steps":["Give the agent a daily schedule trigger (the source uses 6 AM).","Add the Google Calendar Find Event tool to pull today's meetings.","Instruct it to pick out external attendees and research each person and their company with web search.","Add Slack Send Channel Message so it posts one summary paragraph per meeting.","Test in preview mode, then switch it on."],"prompt_template":"Every weekday at [TIME]: find today's events in my calendar. For each attendee whose email domain is not [YOURDOMAIN.COM], research the person and their company on the web.\nPost one Slack message per meeting to #[CHANNEL]: meeting title and time, then 2-3 sentences per external attendee covering role, company and anything recent that is relevant to the meeting.\nIf you can't find reliable information, say so instead of guessing.","watch_outs":["On the free plan, test runs are capped at 10 actions.","The agent may ask for confirmations while testing but runs on its own once live."],"reported_result":"In testing, the agent found correct background on the author's CEO and CTO and posted it to Slack. No results from live runs were reported.","source":{"url":"https://www.xray.tech/post/ai-agent-zapier","site":"XRay.Tech blog","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/research-calendar-guests-and-send-a-slack-briefing","title":"Research calendar guests and send a Slack briefing"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/research-meeting-guests-slack-briefing-zapier-agents"},{"slug":"follow-up-email-drafts-lindy","title":"How to draft follow-up emails automatically when nobody replies (Lindy)","category":"sales-leads","platform":"Lindy","tools":["Lindy","Your email account (sent mail)"],"problem":"Emails that never get an answer are easy to forget, and remembering to nudge each one costs attention.","steps":["Connect a Lindy flow to your sent mail.","After each outgoing email, have it wait a set amount of time.","Check the thread for a reply.","If none arrived, draft a short follow-up.","Deliver the draft to your inbox so you decide whether to send it."],"prompt_template":"Whenever I send an email, wait [3] business days. If the recipient has not replied in the same thread, draft a follow-up of at most 60 words that restates my original request in one line and makes it easy to answer yes or no.\nDo not send it. Save it as a draft and notify me.\nSkip threads where my message was a thank-you, an FYI, or contains \"no reply needed\".","watch_outs":["The follow-up was drafted for review, not sent automatically; keep that until you trust it.","Exclude threads that don't need a follow-up (thank-yous, FYIs)."],"reported_result":"The user had started using it the day before and said it was working well so far. The thread gives no metrics or longer-term results.","source":{"url":"https://www.reddit.com/r/EntrepreneurRideAlong/comments/1awb7xj/has_anyone_here_actually_used_lindy_or_similar_ai","site":"Reddit","published":"2024-02-01"},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/draft-follow-up-emails-automatically-when-no-reply-arrives","title":"Draft follow-up emails automatically when no reply arrives"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/follow-up-email-drafts-lindy"},{"slug":"find-local-business-prospects-ai-agent","title":"How to find local businesses that need a website with an AI research agent","category":"sales-leads","platform":"OpenClaw + ChatGPT/Claude","tools":["ChatGPT or Claude","OpenClaw (agent runner)","Google Places API","An always-on server"],"problem":"Finding local businesses with no website, or a slow one, means hours of clicking through map listings before you can pitch web-design services.","steps":["Choose a city and a business type to target.","Have the agent search Google Maps listings (via the Places API) for that category.","For each listing, check whether it has a website and test that site's speed and mobile experience.","Flag businesses with no site as high priority.","Output a ranked prospect list each day for you to review."],"prompt_template":"Research task only, do not contact anyone.\nFind up to [30] [BUSINESS TYPE] businesses in [CITY] from map listings. For each, record: name, address, phone, rating, number of reviews, and website (or \"none\").\nFor businesses that have a site, note whether it loads well on a phone and whether it looks outdated.\nReturn a table sorted by how strong a prospect each is for [SERVICE]: no website first, then weak websites. Add a one-line reason per row.","watch_outs":["The poster's full setup chained six agents, including email and SMS outreach; check the rules on cold outreach where you operate before going beyond research.","The only stated running cost is a cheap server (about $5-10 a month), plus separately paid APIs such as Google Places."],"reported_result":"The poster says this step produces a daily ranked list without manual searching. The thread gives no conversion, revenue or close-rate numbers, and commenters repeatedly asked for proof that the wider system makes money.","source":{"url":"https://www.reddit.com/r/passive_income/comments/1rmsb2o/i_set_up_6_ai_agents_that_sell_web_services_24_7","site":"Reddit","published":"2026-03-01"},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/find-web-design-prospects-with-sales-agents","title":"Find web design prospects with sales agents"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/find-local-business-prospects-ai-agent"},{"slug":"extract-invoice-data-bookkeeping-sheet-n8n","title":"How to extract invoice data into a bookkeeping sheet with n8n and OpenAI","category":"finance-invoicing","platform":"n8n","tools":["n8n","OpenAI (vision)","Google Drive","Google Sheets","A simple upload web page"],"problem":"Receipts and invoices arrive as photos and PDFs. Typing date, merchant, amount and VAT into a spreadsheet and filing the file is slow and error-prone.","steps":["Build a small upload page that posts the file to an n8n webhook.","Identify the user (the prototype looks up a token in a Google Sheet).","Send images to OpenAI vision; for PDFs, extract the text first and then analyse it.","Have the model return JSON (date, merchant, gross amount, VAT rate, category) and show it to the user to review and edit.","On save, store the file in a Drive folder per user and month, and append the data to a yearly sheet."],"prompt_template":"Extract bookkeeping data from this invoice or receipt. Return JSON only:\n{\"date\": \"YYYY-MM-DD\", \"merchant\": \"...\", \"gross_amount\": number, \"currency\": \"...\", \"vat_rate\": number|null, \"vat_amount\": number|null, \"category\": one of [CATEGORIES], \"invoice_number\": \"...\"|null}\nUse null for anything not clearly printed on the document. Do not calculate or guess missing values.","watch_outs":["The token lookup is a stand-in for real authentication; don't expose it publicly as is.","Keep the human review step: extracted amounts and VAT rates must be checked before they reach your books."],"reported_result":"A working prototype with a demo video. The source gives no accuracy or time-saving figures, and the public repo contains sanitised exports that need your own credentials and IDs.","source":{"url":"https://github.com/Fabiano225/ai-invoice-automation","site":"GitHub","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/extract-invoice-data-and-file-it-into-a-bookkeeping-sheet","title":"Extract invoice data and file it into a bookkeeping sheet"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/extract-invoice-data-bookkeeping-sheet-n8n"},{"slug":"overdue-invoice-reminders-n8n","title":"How to send overdue invoice reminders with an AI agent that adjusts its tone (n8n)","category":"finance-invoicing","platform":"n8n","tools":["n8n (self-hosted in Docker in the source)","Gemini (gemini-2.0-flash-lite)","n8n Data Table","Gmail SMTP"],"problem":"Chasing late payments is awkward. Reminders should be gentle for a few days overdue and firmer after weeks, and nobody wants to write each one.","steps":["Keep invoices in an n8n Data Table: client, email, amount, due date, balance, status.","Create a chat-triggered AI agent with simple memory and two tools: one that reads the invoices table and one that sends email.","In the system prompt, tell it to select invoices that are past due, have a positive balance and are Pending, and to skip ones already Reminded.","Have it soften or firm up the tone depending on how many days overdue each invoice is.","Return a JSON summary of the emails sent."],"prompt_template":"You write payment reminders for [COMPANY]. For each invoice provided, write a short email to the client.\nTone by days overdue: 1-7 friendly reminder; 8-30 polite but clear, restate amount and due date; 31+ firm, mention next steps [NEXT STEPS].\nAlways include invoice number, amount, due date and how to pay: [PAYMENT DETAILS]. Never threaten, never add fees that are not in our terms.\nReturn JSON: [{\"invoice\": \"...\", \"to\": \"...\", \"subject\": \"...\", \"body\": \"...\"}]","watch_outs":["Gmail SMTP sending needs an app password.","The source doesn't update the status to Reminded after sending; add that step or clients will get duplicate reminders.","The agent filters rows itself from a full-table fetch; for larger tables, filter in a normal node first."],"reported_result":"The source describes a finished workflow that runs from a single chat command. It reports no test results or example outputs.","source":{"url":"https://dev.to/canankorkut/build-your-first-ai-agent-with-n8n-32pf","site":"DEV Community","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/send-tone-adjusted-overdue-invoice-reminders","title":"Send tone-adjusted overdue invoice reminders"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/overdue-invoice-reminders-n8n"},{"slug":"generate-invoices-from-template-make","title":"How to generate and email invoices from a template with Make","category":"finance-invoicing","platform":"Make","tools":["Make","Google Sheets","Google Docs (invoice template)","Gmail"],"problem":"Freelancers who send several invoices a week rebuild the same document by hand each time.","steps":["Keep invoice data (client, items, amounts, due date) in a Google Sheet.","Have a Make scenario pick up rows that are ready to invoice.","Fill a Google Docs invoice template with the row's data.","Convert the document to PDF and email it to the client.","Write the invoice status back to the sheet."],"prompt_template":"Optional AI step for the email body:\nWrite a short, friendly email (max 80 words) to [CLIENT NAME] sending invoice [NUMBER] for [AMOUNT] due on [DATE] for [PROJECT/SERVICE]. Mention that the PDF is attached and how to pay: [PAYMENT DETAILS]. No sales pitch, no emojis.","watch_outs":["The source stays high level, so expect to design the template mapping yourself.","Send the first few invoices to yourself to check totals and formatting before going live."],"reported_result":"The poster says it was a game-changer for sending 5-10 invoices a week. The thread gives no hard numbers and keeps setup details private.","source":{"url":"https://www.reddit.com/r/maketemplates/comments/1mqu8s0/stop_wasting_time_on_invoices_i_automated_the","site":"Reddit","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/generate-invoices-automatically-from-a-template","title":"Generate invoices automatically from a template"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/generate-invoices-from-template-make"},{"slug":"supplier-invoice-approval-pipeline-n8n","title":"How to set up a supplier-invoice approval pipeline from Gmail with n8n","category":"finance-invoicing","platform":"n8n","tools":["n8n (queue mode)","PostgreSQL","Redis","Docker","Gmail","OCR","OpenAI","Slack"],"problem":"Supplier invoices arrive by email and each one needs data extraction, duplicate and supplier checks, approval for larger amounts and an accounting entry.","steps":["Run n8n in queue mode with PostgreSQL, Redis and a worker in Docker.","Create tables for suppliers, invoices and an audit log, and add Gmail, OpenAI, Postgres and Slack credentials.","Plan the flow: Gmail, then OCR, then AI extraction.","Then duplicate and supplier checks, classification, and approval routing for high-value invoices.","Finally create the journal entry and schedule the payment."],"prompt_template":"Classify this supplier invoice for accounts payable. Return JSON:\n{\"supplier_name\": \"...\", \"supplier_vat_id\": \"...\"|null, \"invoice_number\": \"...\", \"invoice_date\": \"YYYY-MM-DD\", \"due_date\": \"YYYY-MM-DD\"|null, \"net\": number, \"tax\": number|null, \"gross\": number, \"currency\": \"...\", \"expense_category\": one of [CATEGORIES], \"needs_approval\": true if gross > [THRESHOLD]}\nUse null when a value is not on the document.\n\nText: {{ocr_text}}","watch_outs":["Queue mode can retry stalled jobs, so every step must be idempotent; duplicate detection is essential.","Back up the n8n encryption key, or you lose access to stored credentials."],"reported_result":"This is part 1 of a three-part series and covers the infrastructure only. It ends with the stack and schema running; extraction, validation, approval and accounting are left to later parts. No results yet.","source":{"url":"https://medium.com/@beulahdaizy/i-built-an-ai-powered-invoice-automation-pipeline-in-n8n-heres-how-part-1-the-stack-602d839eaf41","site":"Medium","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/process-supplier-invoices-from-gmail-through-approval-and-pa","title":"Process supplier invoices from Gmail through approval and payment"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/supplier-invoice-approval-pipeline-n8n"},{"slug":"weekly-payment-digest-email-n8n","title":"How to turn payment notification emails into one weekly digest with n8n","category":"finance-invoicing","platform":"n8n","tools":["n8n","Gmail (with a label)","OpenRouter (gpt-4o-mini)","n8n Code node"],"problem":"Payment and transaction notifications arrive as a stream of separate emails (about 15 a week for the author), so it is hard to see the week at a glance.","steps":["Create a Gmail label (or filter) for payment emails.","Schedule the n8n workflow for Sunday at 8 AM.","Fetch the past week's labelled messages, aggregate them and truncate the combined text to 100K characters.","Have an AI Agent node write the digest from a system-message template.","Convert the Markdown to HTML in a Code node and send it with Gmail."],"prompt_template":"You summarise a week of payment notification emails for [NAME].\nProduce: 1) a one-line total per currency, 2) a table of payments (date, payer or merchant, amount, currency, what it was for), 3) anything unusual: duplicates, failed payments, unknown merchants.\nOnly use amounts that appear in the emails. Format as Markdown.\n\nEmails: {{emails}}","watch_outs":["Truncating to 100K characters keeps the input within the model's limits.","Without the Markdown-to-HTML step, the email shows raw Markdown symbols."],"reported_result":"The workflow turns about 15 weekly notifications into a single digest email. The post shows a successful test run but no further measurements.","source":{"url":"https://mariannanakos.substack.com/p/n8n-tutorial-built-an-ai-workflow","site":"Substack","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/email-a-weekly-digest-of-payment-notification-emails","title":"Email a weekly digest of payment notification emails"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/weekly-payment-digest-email-n8n"},{"slug":"reconcile-bank-statement-invoices-claude-cowork","title":"How to reconcile a bank statement with invoices using Claude Cowork","category":"finance-invoicing","platform":"Claude Cowork","tools":["Claude Cowork","A local folder with your bank statement and invoices","Excel"],"problem":"Matching each bank transaction to an invoice every month, cleaning vendor names and spotting missing receipts takes an afternoon.","steps":["Put the month's bank statement and the related invoices in one folder.","Give Cowork access to that folder.","In one prompt, ask it to clean vendor names, categorise expenses and match invoices to transactions.","Ask it to flag transactions with no invoice and produce a formatted Excel file.","Have it rename the invoice files to a consistent pattern."],"prompt_template":"In this folder there is a bank statement for [MONTH] and invoice files.\n1. Read every transaction. Clean up vendor names (e.g. \"AMZN MKTP DE*2X\" becomes \"Amazon\").\n2. Categorise each one using: [CATEGORIES].\n3. Match invoices to transactions by amount, date (+/- 5 days) and vendor.\n4. Create reconciliation-[MONTH].xlsx with columns: date, vendor, amount, category, matched invoice file, status (matched / missing invoice / unsure).\n5. Rename matched invoices to YYYY-MM-DD_vendor_amount.pdf.\nDo not delete or move any other files. List anything you were unsure about at the end.","watch_outs":["Start with a small, non-sensitive folder: Cowork reads your local files directly.","Spot-check the matches before handing the file to your accountant."],"reported_result":"The author says it extracted everything correctly, renamed the invoices cleanly and marked every transaction without an invoice, saving her a full afternoon a month. She only tested a few invoices.","source":{"url":"https://aiblewmymind.substack.com/p/claude-cowork-use-cases-guide","site":"Substack","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/reconcile-bank-statements-with-invoices","title":"Reconcile bank statements with invoices"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/reconcile-bank-statement-invoices-claude-cowork"},{"slug":"find-recurring-subscriptions-bank-statements-claude-cowork","title":"How to find forgotten subscriptions in your bank statements with Claude Cowork","category":"finance-invoicing","platform":"Claude Cowork","tools":["Claude Cowork","Bank statements (CSV, XLS, XLSX or PDF)","Optional: Windows Task Scheduler"],"problem":"Small recurring charges are easy to forget. The author's example was a €5/month subscription noticed seven months late.","steps":["Save statements in a folder, one subfolder per bank, with the month in each filename.","Ask Cowork to read every file, whatever the format.","Have it flag charges that repeat across two or more months and leave out one-off purchases, groceries and bank fees.","Ask for a dashboard with one card per subscription and an .xlsx report saved to the folder.","Optionally rerun the prompt monthly."],"prompt_template":"Read all bank statements in this folder (subfolders per bank, month in each filename).\nFind charges that recur in at least two different months from the same merchant at a similar amount. Exclude groceries, one-off purchases, transfers and bank fees.\nFor each subscription list: merchant, typical amount, frequency, first and last seen, bank.\nSave subscriptions.xlsx in this folder with those columns plus an empty \"Keep / Cancel\" column. Don't cancel anything.","watch_outs":["The source suggests scheduling the prompt monthly with Task Scheduler.","A 'Canceled' column stays empty unless you confirm cancellations yourself."],"reported_result":"The source shows the dashboard and report as screenshots but gives no counts or savings totals.","source":{"url":"https://artificialcorner.com/p/cowork-workflows","site":"Artificial Corner (Substack)","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/find-recurring-subscriptions-in-bank-statements","title":"Find recurring subscriptions in bank statements"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/find-recurring-subscriptions-bank-statements-claude-cowork"},{"slug":"track-subscriptions-renewals-gmail-chatgpt-agent","title":"How to list subscriptions and renewal dates from Gmail with ChatGPT Agent","category":"finance-invoicing","platform":"ChatGPT Agent","tools":["ChatGPT Agent mode","Gmail access"],"problem":"Renewal notices are scattered through years of email, so it is hard to know what you pay for and when it renews.","steps":["Turn on Agent mode in ChatGPT and connect Gmail.","Ask it to search for terms like renewal, subscription and auto-renew, or for specific service names.","Have it open each relevant email and extract service, renewal date and cost.","Ask for a short list you can review.","Tell it to ask permission before unsubscribing or cancelling anything."],"prompt_template":"Search my Gmail for subscription and renewal emails from the last [12] months (terms: renewal, subscription, auto-renew, \"your plan\", receipt, plus these services: [LIST]).\nFor each service, give: name, cost, billing period, next renewal date (if stated), and the email it came from.\nShow the list sorted by next renewal date. Do not cancel, unsubscribe or reply to anything without asking me first.","watch_outs":["This requires giving ChatGPT access to your email, which won't suit everyone; a plain cancellation reminder is the low-tech alternative."],"reported_result":"The writer says it helped her cancel unused streaming services. The article gives no figures and doesn't show the resulting list.","source":{"url":"https://www.tomsguide.com/ai/5-clever-ways-i-use-chatgpt-agent-to-save-serious-money-each-month","site":"Tom's Guide","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/track-subscriptions-and-renewals-from-gmail","title":"Track subscriptions and renewals from Gmail"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/track-subscriptions-renewals-gmail-chatgpt-agent"},{"slug":"email-triage-agent-make","title":"How to triage unread email with Make's AI Agent and get a morning summary","category":"inbox-calendar","platform":"Make","tools":["Make (AI Agent module)","Gmail","OpenAI API key"],"problem":"Waking up to a full inbox means sorting newsletters, follow-ups and urgent items before doing any real work.","steps":["Start a Make scenario with Gmail's Watch New Emails trigger for unread mail.","Pass each email to Make's AI Agent module running on your OpenAI key.","Give it your triage rules for sender, subject and content.","Have it label, archive or flag each email.","Send yourself a summary email of what it did."],"prompt_template":"You triage my inbox. For each email decide one action:\n- URGENT: needs my reply today (clients [LIST], anything about [TOPICS])\n- FOLLOW-UP: needs action this week\n- READ-LATER: newsletters and updates I subscribed to\n- ARCHIVE: notifications, receipts, promotions\nReturn JSON {\"action\": \"...\", \"label\": \"...\", \"one_line_summary\": \"...\"}. Never delete. When unsure, choose FOLLOW-UP.\n\nFrom: {{from}}\nSubject: {{subject}}\nBody: {{body}}","watch_outs":["The author puts the cost at about $12-15 a month: a paid Make plan from $9 plus $5-10 of OpenAI credits lasting 1-3 months.","Prefer labelling over archiving until you trust the rules."],"reported_result":"The author says he woke up to a summary of 3 urgent emails, 2 newsletter ideas, 1 follow-up and 4 archived emails without opening Gmail. The system prompt sits behind a paywall, and the source doesn't explain how the approval step works.","source":{"url":"https://aimaker.substack.com/p/build-ai-email-triage-agent-automation-make-tutorial","site":"Substack","published":"2026-01-15"},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/triage-unread-email-with-an-approval-step","title":"Triage unread email with an approval step"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/email-triage-agent-make"},{"slug":"booking-emails-to-calendar-events-n8n","title":"How to turn booking confirmation emails into calendar events with n8n","category":"inbox-calendar","platform":"n8n","tools":["n8n","Gmail (label)","PDF parsing","LLM extractor","Google Calendar"],"problem":"Travel and booking confirmations pile up in the inbox (the source built this for a CEO) and someone has to copy each one into the calendar.","steps":["Watch Gmail for messages labelled \"Events\".","Wrap the email body into a PDF so it goes through the same extractor as any PDF attachments.","Extract date, time, location, confirmation number, vendor and notes from both sources in parallel.","Merge the results, with attachment data taking priority.","Create an event in a separate \"Auto-imported\" calendar with a link to the email, or flag the email and send a missing-fields notice."],"prompt_template":"Extract calendar details from this booking confirmation. Return JSON:\n{\"title\": \"...\", \"start\": \"YYYY-MM-DDTHH:MM\", \"end\": \"...\"|null, \"timezone\": \"...\"|null, \"location\": \"...\"|null, \"confirmation_number\": \"...\"|null, \"vendor\": \"...\", \"notes\": \"...\", \"missing_fields\": [list]}\nIf there are several separate bookings, return an array. Never guess a date or time that is not written in the text.","watch_outs":["A person still has to apply the \"Events\" label to trigger the workflow.","Use a separate calendar so wrong imports are easy to spot and delete."],"reported_result":"Events are now created automatically in a dedicated calendar instead of being entered by hand. The author notes there's no step yet for confirmations that contain several events.","source":{"url":"https://www.reddit.com/r/n8n/comments/1tc4fpb/built_an_n8n_workflow_that_turns_any_booking","site":"Reddit","published":"2026-05-13"},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/turn-booking-confirmation-emails-into-calendar-events","title":"Turn booking confirmation emails into calendar events"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/booking-emails-to-calendar-events-n8n"},{"slug":"bulk-unsubscribe-newsletters-claude","title":"How to bulk-unsubscribe from Gmail newsletters with Claude Cowork and Claude in Chrome","category":"inbox-calendar","platform":"Claude Cowork","tools":["Claude Cowork","Gmail connector","Claude in Chrome extension","Paid Claude plan (Pro in the source)"],"problem":"Dozens of newsletters you no longer read keep arriving, and unsubscribing one by one is a chore.","steps":["Create a Cowork project with the Gmail connector enabled.","Point Claude at a label or search that holds the newsletters (the author used an \"archivable\" label).","Ask it to list newsletters that look safe to drop, without acting yet.","Go through its list and decide keep or cut for each.","Let it unsubscribe from the ones you cut via the Chrome extension."],"prompt_template":"Look at emails under the label [LABEL] from the last [3] months. List every newsletter or marketing sender with: sender, how many emails, and your recommendation keep/cut with a short reason.\nDo not unsubscribe or change anything yet. Wait for me to confirm which ones to cut. After I confirm, unsubscribe only from those senders and report what you did.","watch_outs":["The author explicitly told Claude to list candidates first and not act yet; keep that order.","Cowork requires a paid Claude plan (the author used Pro)."],"reported_result":"The PCWorld writer reports Claude unsubscribed from 21 newsletters in about five minutes, one for each he chose to cut, and suggested marking any that come back as spam.","source":{"url":"https://www.pcworld.com/article/3185314/i-was-buried-in-email-newsletters-heres-how-claude-rescued-me.html","site":"PCWorld","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/bulk-unsubscribe-from-gmail-newsletters","title":"Bulk-unsubscribe from Gmail newsletters"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/bulk-unsubscribe-newsletters-claude"},{"slug":"scheduled-claude-morning-briefing","title":"How to run Claude on a schedule for a personal morning briefing","category":"inbox-calendar","platform":"Claude Code (claude -p)","tools":["Claude Code CLI (claude -p)","cron","A personal context / memory folder","Optional: email, Telegram or voice-text triggers"],"problem":"You want a short, personal briefing every morning without opening a dozen apps, and a way to hand small tasks to an assistant by message.","steps":["Write a context file (or folder) about yourself, your priorities and ongoing projects, plus a place for long-term notes.","Schedule a cron job that runs claude -p with a briefing prompt in the early morning.","Let the run read your context and memory and write the briefing (the author also generates a personalised podcast).","Optionally let incoming emails, voice texts or Telegram messages start a session and get a reply.","Read the briefing in the morning and update the context as things change."],"prompt_template":"It's [DATE], early morning. Read context/me.md, context/projects.md and memory/notes.md.\nWrite my briefing in under 250 words: 1) the 3 things that matter most today and why, 2) deadlines in the next 7 days, 3) anything in memory I said I'd follow up on, 4) one suggestion I might not have thought of.\nAppend new facts worth remembering to memory/notes.md. Do not send emails or messages.","watch_outs":["The value depends on rich personal context and memory; a bare scheduled prompt won't do the same.","Keep the scheduled job read-only unless you have explicit approval steps for actions."],"reported_result":"The source gives no numbers. The author says that once it has enough context, it sometimes surfaces useful, even surprising suggestions overnight.","source":{"url":"https://news.ycombinator.com/item?id=47724491","site":"Hacker News","published":"2026-04-10"},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/run-claude-on-a-cronjob-for-morning-briefings-and-email-task","title":"Run Claude on a cronjob for morning briefings and email tasks"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/scheduled-claude-morning-briefing"},{"slug":"newsletters-to-notion-summaries-n8n","title":"How to summarise newsletters into a Notion database with n8n","category":"research-monitoring","platform":"n8n","tools":["n8n","IMAP email","OpenAI (summarisation chain)","Notion"],"problem":"Useful newsletters go unread because there are too many of them; you want the key points, with links, in one searchable place.","steps":["Read unread emails over IMAP in n8n.","Filter by date and by a list of allowed senders.","Summarise each remaining newsletter into bullet points that keep the source URLs.","Split summaries that exceed Notion's text-block limit (the source splits at 1,950 characters) with a Code node.","Create the Notion page from the first chunk and append the rest to it."],"prompt_template":"Summarise this newsletter for my research notes.\n- 3 to 7 bullet points, each one fact or idea, no hype.\n- Keep the original link next to each bullet it came from.\n- End with one line: \"Worth reading in full?\" yes/no and why.\n\nNewsletter: {{email_text}}","watch_outs":["Only OpenAI ran reliably for the author; Claude hit token limits mid-run after tokens were spent, and a local Llama 2 errored.","The author planned to switch from map-reduce to the simpler 'stuff' summarisation method."],"reported_result":"In the walkthrough, 15 unread emails were narrowed to 3 summaries, all written to Notion. The IMAP filter rules were unreliable and the author first picked the wrong summarisation method.","source":{"url":"https://medium.com/@syrom_85473/a-practical-n8n-workflow-example-from-a-to-z-part-2-aggregating-newsletter-summaries-in-a-688f30f69352","site":"Medium","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/summarize-newsletters-into-a-notion-database","title":"Summarize newsletters into a Notion database"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/newsletters-to-notion-summaries-n8n"},{"slug":"daily-arxiv-papers-to-notion-n8n","title":"How to track new arXiv papers in Notion every day with n8n and Gemini","category":"research-monitoring","platform":"n8n","tools":["n8n","arXiv API","Gemini 2.5 Flash","Notion","Optional: Gmail, Feishu"],"problem":"Keeping up with new papers on a topic means checking arXiv every day and skimming abstracts.","steps":["Schedule the workflow daily (6:00 AM in the source).","Query the arXiv API for the previous day's submissions on your keyword.","Clean the Atom response into JSON, with authors and categories as arrays.","Have Gemini analyse each paper and write a daily summary.","Create pages in Notion and optionally send an email digest."],"prompt_template":"For each arXiv paper below, write: a one-sentence plain-English summary, the main contribution, the method, and who should read it. Then write a daily overview (max 120 words) of the trends across all papers on [TOPIC].\nOnly use the title and abstract given; don't speculate about results that are not stated.\n\nPapers: {{papers_json}}","watch_outs":["Notion's create-page node only adds entries and never updates them.","Multi-select fields need real arrays, and null values cause a 400 error."],"reported_result":"The workflow template gives no measured results; it expects fewer than 20 papers a day.","source":{"url":"https://n8n.io/workflows/8847-daily-rag-research-paper-hub-with-arxiv-gemini-ai-and-notion/","site":"n8n workflow library","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/track-daily-arxiv-papers-into-a-notion-database","title":"Track daily arXiv papers into a Notion database"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/daily-arxiv-papers-to-notion-n8n"},{"slug":"url-to-structured-notion-note-n8n","title":"How to turn any article URL into a structured Notion note with n8n","category":"research-monitoring","platform":"n8n","tools":["n8n","OpenAI (JSON mode)","Notion"],"problem":"Saving articles as bare links means you never find or reread them; you want a summary, key insights and tags stored with each one.","steps":["Receive a URL on an n8n webhook.","Fetch the page with an HTTP Request node.","Strip scripts, styles and tags in a Code node and truncate the text (8,000 characters in the source).","Ask OpenAI in JSON mode for title, summary, key insights and tags.","Map the fields straight into Notion database properties."],"prompt_template":"From the article text below, return JSON only:\n{\"title\": \"...\", \"summary\": \"2-3 sentences\", \"key_insights\": [\"...\", \"...\", \"...\"], \"tags\": [up to 5 lowercase tags, reuse these if they fit: TAG LIST], \"content_type\": \"tutorial|opinion|news|research|other\"}\nIf the text is empty or looks like a cookie wall, return {\"error\": \"no content\"}.\n\nArticle: {{text}}","watch_outs":["Share the Notion database with your n8n integration first, or writes fail silently.","JavaScript-heavy single-page apps return empty HTML; a Mozilla/5.0 User-Agent header clears most 403 errors."],"reported_result":"The author reports that stripping HTML first cut the payload from about 60K tokens to under 4K, which improved summaries and cut cost significantly.","source":{"url":"https://community.n8n.io/t/turns-any-article-into-a-structured-notion-note/297566","site":"n8n Community forum","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/turn-any-url-into-a-structured-notion-note","title":"Turn any URL into a structured Notion note"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/url-to-structured-notion-note-n8n"},{"slug":"telegram-second-brain-links-n8n","title":"How to build a Telegram second brain for saving and searching links (n8n)","category":"research-monitoring","platform":"n8n","tools":["n8n","Telegram bot","Jina Reader","Google Sheets","Supabase (free tier)","Gemini embeddings"],"problem":"Links you send yourself in chats get lost. You want to save them in one tap and later ask questions about what you saved.","steps":["Create a Telegram AI agent that sorts each message: a URL is saved, a question is searched, anything else gets a short how-to reply.","For saving, call a webhook that normalises the URL and reads the page with Jina Reader.","Have an AI 'librarian' write a summary and reuse existing categories.","Upsert the entry into Google Sheets by URL and embed the full text into Supabase.","Give the agent two search tools: one over the Sheet, one over the page text."],"prompt_template":"You are my link librarian.\nWhen I send a URL: summarise the page in 2 sentences, pick 1-2 categories from this list [EXISTING CATEGORIES] (create a new one only if nothing fits), and save it.\nWhen I ask a question: search my saved links and answer in 3 bullets with the matching links. If nothing matches, say so.\nOtherwise: reply with one line explaining how to use you.","watch_outs":["Setup needs a Telegram bot token, Supabase, Gemini embeddings, a shared header-auth credential and your own webhook URL in the save tool."],"reported_result":"The author says it works: resending a link updates the existing entry instead of duplicating it, and unreadable pages aren't filed. No usage numbers were shared.","source":{"url":"https://community.n8n.io/t/second-brain-on-telegram-save-and-search-your-links-google-sheets-supabase-rag/304330","site":"n8n Community forum","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/save-and-search-links-via-a-telegram-second-brain","title":"Save and search links via a Telegram second brain"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/telegram-second-brain-links-n8n"},{"slug":"youtube-latest-video-summary-google-doc-make","title":"How to summarise a YouTube channel's latest video into a Google Doc daily with Make","category":"research-monitoring","platform":"Make","tools":["Make","YouTube module","Hugging Face (summarisation)","Google Docs"],"problem":"Following a channel you like means checking for new uploads every day just to see whether they are worth watching.","steps":["Add a Make scheduler that fires daily (9 AM in the source).","Use the YouTube module to look up the channel and get the latest video's title and description.","Send that text to a Hugging Face summarisation model with a set summary length.","Create or find a Google Doc.","Append the summary with the date and video link."],"prompt_template":"Summarise this new video from [CHANNEL] in 3 bullet points based only on the title and description. Add one line: \"Watch if you care about: ...\". If the description is too short to summarise, say \"Description too short, check the video.\"\n\nTitle: {{title}}\nDescription: {{description}}","watch_outs":["The summariser only sees the title and description, not a transcript, so quality depends on how detailed the description is."],"reported_result":"The author says the daily check-and-copy work no longer needs doing by hand. No measured results; the summary only covers the title and description.","source":{"url":"https://www.labellerr.com/blog/building-ai-agents-with-make-com-a-no-code-guide/","site":"Labellerr blog","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/summarize-a-youtuber-s-latest-video-into-a-google-doc-daily","title":"Summarize a YouTuber's latest video into a Google Doc daily"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/youtube-latest-video-summary-google-doc-make"},{"slug":"newsletter-briefing-podcast-claude-cowork-n8n","title":"How to build a daily newsletter briefing with Claude Cowork and an n8n MCP bridge","category":"research-monitoring","platform":"Claude Cowork + n8n","tools":["Claude Cowork","n8n (as an MCP server)","Readwise","Word","Optional: a cloud text-to-speech service"],"problem":"You subscribe to many newsletters and want a short daily briefing, ideally as audio, instead of reading them all.","steps":["Forward your newsletters to Readwise.","Run n8n as an MCP server with a workflow that calls the Readwise API (Cowork had no direct network access in the source).","Have Cowork fetch the previous day's articles through that bridge.","Ask it for a two-page Word briefing with an executive summary.","For audio, use a proper text-to-speech service rather than a local fallback."],"prompt_template":"Using the articles fetched from Readwise for [DATE], write a briefing of at most two pages:\n1. Executive summary: 5 bullets.\n2. One section per theme, with the article titles and links it draws on.\n3. \"Skip list\": items that look important but aren't, one line each.\nWrite in plain, spoken-friendly sentences so it can be read aloud.","watch_outs":["Cowork couldn't reach cloud TTS models; the author planned to rebuild it in Claude Code with ElevenLabs or OpenAI TTS."],"reported_result":"The briefing document worked and the n8n bridge ran without problems. The audio step failed: without access to a good TTS model, Cowork used a local library and the result was too robotic to listen to.","source":{"url":"https://ishyshkova.substack.com/p/claude-cowork-review-real-use-cases","site":"Substack","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/generate-a-daily-podcast-briefing-from-newsletters","title":"Generate a daily podcast briefing from newsletters"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/newsletter-briefing-podcast-claude-cowork-n8n"},{"slug":"reddit-threads-to-notion-insights-n8n","title":"How to mine Reddit threads into Notion insight reports with n8n and Gemini","category":"research-monitoring","platform":"n8n","tools":["n8n","Reddit RSS","Bright Data","Gemini","Notion"],"problem":"Marketers want to know what customers complain about and ask for in a subreddit, without reading hundreds of threads.","steps":["Schedule the workflow for a chosen subreddit.","Read the subreddit's RSS feed for recent post URLs.","Send the URLs to Bright Data in a batch and poll (every 30 seconds in the source) until the snapshot is ready.","Download titles, comments and upvotes and have a Gemini agent extract concerns, themes and marketing insights.","Convert the output into Notion blocks on a new page."],"prompt_template":"You analyse Reddit discussions for [BRAND / PRODUCT CATEGORY]. From the posts and comments below, report:\n1. Top 5 recurring problems or complaints, with links to the threads.\n2. Top 5 things people wish existed.\n3. Words and phrases people use to describe the problem (for copywriting).\n4. Brands mentioned, with sentiment.\nQuote at most one short phrase per point. Don't invent numbers.\n\nData: {{threads_json}}","watch_outs":["Converting the agent's Markdown into valid Notion blocks was the hardest part.","Respect Reddit's and Bright Data's terms for the data you collect."],"reported_result":"The author demoed it on a skincare subreddit and says it runs end to end without manual steps. No volume or accuracy numbers; multi-subreddit support was still future work.","source":{"url":"https://dev.to/kichuman28/unstoppable-reddit-research-agent-real-time-insights-with-n8n-bright-data-34g0","site":"DEV Community","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/mine-reddit-threads-into-structured-notion-reports","title":"Mine Reddit threads into structured Notion reports"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/reddit-threads-to-notion-insights-n8n"},{"slug":"weekly-competitor-monitoring-brief-claude","title":"How to get a weekly competitor-monitoring brief from Claude","category":"research-monitoring","platform":"Claude","tools":["Claude (scheduled task)","Web access","Your competitor list"],"problem":"Checking competitors' sites for pricing changes, launches and hiring signals is a weekly chore that usually gets skipped.","steps":["Write down the competitors you want to track.","Schedule Claude to run every Monday morning (8 AM in the source).","Have it check each site for pricing changes, new products or features, new content, announcements and job listings.","Ask it to summarise what changed across all competitors.","Have it highlight the single most important item of the week."],"prompt_template":"Every Monday: for each company in [COMPETITOR LIST], check their website, pricing page, blog or changelog, newsroom and careers page.\nReport only what changed in the last 7 days: pricing, new products or features, notable content, announcements, and new job openings that hint at strategy (say what they might signal).\nInclude a link for every item. If nothing changed, say \"no changes found\".\nEnd with \"Most important this week:\" one item and why it matters to [MY COMPANY].","watch_outs":["The only concrete setup detail in the source is the schedule, so test the prompt for a few weeks before relying on it.","Ask for a link for every claimed change so you can verify it."],"reported_result":"The poster gets a Monday brief without manual checking or a dashboard, and uses job listings to infer likely moves (such as a push on distribution). No measured accuracy, savings or examples of misses.","source":{"url":"https://www.reddit.com/r/PromptEngineering/comments/1tz5f6n/i_set_claude_up_as_an_agent_that_spies_on_my","site":"Reddit","published":"2025-06-05"},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/run-a-monday-competitor-monitoring-brief-from-job-listings","title":"Run a Monday competitor-monitoring brief from job listings"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/weekly-competitor-monitoring-brief-claude"},{"slug":"rss-to-bluesky-posts-zapier-agents","title":"How to auto-post blog summaries from RSS to Bluesky with Zapier Agents","category":"content-social","platform":"Zapier Agents","tools":["Zapier Agents (Zapier Central in the source)","RSS by Zapier","Bluesky","A style guide document"],"problem":"Every new blog post should be promoted on social media, but writing a hook for each one is easy to skip.","steps":["Create an agent with an RSS by Zapier trigger on your blog's feed.","Connect Bluesky's Create Post action.","Upload a style guide as a knowledge source.","Instruct the agent to write a short, punchy hook that links to the article, and let it fill the action fields.","Test in preview, optionally turn on Require Approval, then publish."],"prompt_template":"A new article was published: {{title}} ({{link}})\n{{summary}}\nWrite one Bluesky post (max 280 characters) following the attached style guide: open with a specific, curiosity-driving hook drawn from the article, no clickbait, at most one emoji, no hashtags, end with the link.","watch_outs":["The free plan allows 400 activities a month.","Turn on Require Approval to review each post before it goes out."],"reported_result":"The author's agent is live and keeps Bluesky stocked with hooks for new articles while they're away. No post counts or engagement figures.","source":{"url":"https://medium.com/@heelextension/building-your-first-ai-agent-in-zapier-central-aeb2cd0adca0","site":"Medium","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/auto-post-blog-summaries-to-bluesky-from-rss","title":"Auto-post blog summaries to Bluesky from RSS"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/rss-to-bluesky-posts-zapier-agents"},{"slug":"social-posts-from-google-sheet-make","title":"How to write, design and publish social posts from a Google Sheet with Make","category":"content-social","platform":"Make","tools":["Make","Google Sheets","OpenAI","An image model (DALL·E or similar)","Cloudinary","Slack","Buffer"],"problem":"Producing a caption and an image for every post, then scheduling it, takes much longer than coming up with the idea.","steps":["Add a Google Sheets row with the post idea, audience and tone.","In Make, have OpenAI write the caption.","Generate an image with an image model and upload it to Cloudinary.","Write the assets back to the sheet and send a Slack approval request.","After approval in Slack, publish via Buffer."],"prompt_template":"Write a [PLATFORM] caption for [BRAND].\nIdea: {{idea}}\nAudience: {{audience}}\nTone: {{tone}}\nRules: the first line must work on its own; max [150] words; one call to action; max 3 relevant hashtags at the end; no claims we can't back up.\nAlso return a one-sentence image prompt that matches the caption, in this style: [STYLE NOTES].","watch_outs":["It is not fully hands-off: approval happens in Slack before anything is published.","The builder says adding error handling and filters was what made the Make scenario stable."],"reported_result":"The poster says a sheet row becomes a published post in under two minutes and calls it consistent, while still keeping a human approval step.","source":{"url":"https://www.reddit.com/r/automation/comments/1lff4ph/i_built_an_automation_that_writes_designs_and","site":"Reddit","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/write-design-and-post-social-content-from-a-google-sheet","title":"Write, design and post social content from a Google Sheet"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/social-posts-from-google-sheet-make"},{"slug":"repurpose-newsletter-into-social-content-claude-cowork","title":"How to repurpose a newsletter into carousel, reel and LinkedIn content with Claude Cowork","category":"content-social","platform":"Claude Cowork","tools":["Claude Cowork","Per-platform skills","Examples of your best past posts","A voice guide"],"problem":"Each newsletter could feed a week of social posts, but rewriting it for every platform takes hours.","steps":["Write a voice guide and collect reference examples of your best posts per platform.","Create a repurposing skill for each format (carousel, reel script, LinkedIn post, Substack note).","When a newsletter is finished, give it to Cowork with a single prompt.","Review the generated pieces and adjust them.","Refine the skills and voice guide over time."],"prompt_template":"Here is this week's newsletter: [PASTE OR FILE].\nUsing my voice guide (voice.md) and the examples in /examples, create:\n1. A 7-slide carousel (headline plus max 25 words per slide).\n2. A 45-second reel script with on-screen text cues.\n3. A LinkedIn post (max 200 words, hook in the first line).\n4. A Substack note (max 60 words).\nKeep every claim faithful to the newsletter. Don't add statistics that aren't in it.","watch_outs":["Setup takes real time: without the skills, references and voice guide she says the output turns generic."],"reported_result":"The creator says repurposing now saves her a great deal of time, though she gives no figures.","source":{"url":"https://taylaburrell.substack.com/p/13-actually-useful-things-you-can","site":"Substack","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/repurpose-a-newsletter-into-social-content","title":"Repurpose a newsletter into social content"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/repurpose-newsletter-into-social-content-claude-cowork"},{"slug":"clickup-daily-digest-email-make","title":"How to email daily ClickUp task digests to each project's stakeholders with Make","category":"team-ops","platform":"Make","tools":["Make","ClickUp","Email"],"problem":"Stakeholders want a daily view of open action items for their project, without logging into ClickUp.","steps":["In ClickUp, keep a master action-item list linked to project and contact lists so each task carries its stakeholders' emails.","Schedule a Make scenario (6 AM in the source) that lists incomplete tasks.","Iterate over the tasks and group them by project with a Text Aggregator, mapping individual fields.","Send one digest email per project to that project's stakeholder email field.","Test outputs and email format before switching on live sends."],"prompt_template":"Optional AI step to make the digest readable:\nTurn this list of open tasks for project [PROJECT] into a short email: a one-line status, then tasks grouped by owner with due dates, overdue items first. Max 200 words. Use only the tasks provided.\n\nTasks: {{tasks}}","watch_outs":["Write the generated subject, recipients and body to Google Sheets or a Make data store first, and only enable real emails once they look right."],"reported_result":"The first version failed: the aggregator pulled whole ClickUp objects, so the email showed raw JSON. Commenters said the fix was mapping individual fields and grouping by project name. No finished result was reported.","source":{"url":"https://www.reddit.com/r/Make/comments/1u6amg9/clickup_x_makecom_automation_my_boss_wants_me_to","site":"Reddit","published":"2026-07-01"},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/email-daily-clickup-digests-to-stakeholders","title":"Email daily ClickUp digests to stakeholders"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/clickup-daily-digest-email-make"},{"slug":"async-standup-summaries-zapier-agents","title":"How to replace daily standups with async updates and AI summaries (Zapier Agents)","category":"team-ops","platform":"Zapier Agents","tools":["Zapier Agents","Slack"],"problem":"Daily standups take a lot of team time for information that could be written down in a minute.","steps":["Build one agent that pings each team member in Slack for an async update.","Build a second agent that collects the replies and summarises them for everyone.","Post the summary to the team channel.","Test both agents in a sandbox channel first.","Once the signal is good, move them to the real team channel and cut back live meetings."],"prompt_template":"Ask (DM to each person at [TIME]): \"Quick async standup: 1) what you finished, 2) what you're on today, 3) anything blocking you?\"\nSummary (at [TIME]): combine all replies into one Slack post: blockers first (with who can help), then one line per person. Mark anyone who didn't reply. Don't add anything that wasn't said.","watch_outs":["Launch in a sandbox channel and check the signal-to-noise ratio before rolling out."],"reported_result":"According to the source, Zapier's team turned its daily standup into a twice-a-week sync almost overnight. No figures on time saved or summary quality.","source":{"url":"https://getdx.com/blog/using-ai-to-accelerate-hiring-and-productivity-at-zapier/","site":"DX blog","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/collect-and-summarize-async-standup-updates","title":"Collect and summarize async standup updates"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/async-standup-summaries-zapier-agents"},{"slug":"sprint-changelog-from-linear-claude-cowork","title":"How to write end-of-sprint changelogs from completed Linear tickets with Claude Cowork","category":"team-ops","platform":"Claude Cowork","tools":["Claude Cowork (recurring task)","Linear (via MCP)","Your changelog / publishing tool (via MCP)"],"problem":"Writing release notes every sprint means rereading every finished ticket and translating it into user-facing language.","steps":["Connect Claude to Linear and to your changelog destination through MCP.","Create a recurring Cowork task for the end of each sprint (every two weeks in the source).","Have Claude read completed issues and decide which ones are user-facing.","Ask it to draft changelog copy from the ticket descriptions and comments.","Define rules for when an update also deserves an email or in-app announcement, and review before publishing."],"prompt_template":"Read all Linear issues completed in the sprint ending [DATE] for team [TEAM].\n1. Keep only user-facing changes; list internal ones separately so I can check.\n2. Group them as New, Improved, Fixed. One line each, written for customers, no ticket jargon.\n3. Recommend whether this release deserves an email or in-app announcement, using this rule: [RULE].\nSave as a draft. Do not publish until I approve.","watch_outs":["You need two MCP connections (Linear and the publishing destination) plus explicit rules for bigger announcements."],"reported_result":"The PM says it replaced about an hour of manual work per sprint and that the writing is better because Claude includes details they used to skip. They still review it before shipping.","source":{"url":"https://www.reddit.com/r/ClaudeAI/comments/1rvw4jg/cowork_replaced_an_hour_of_my_most_hated_pm_task","site":"Reddit","published":"2026-04-05"},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/write-end-of-sprint-changelogs-from-completed-tickets","title":"Write end-of-sprint changelogs from completed tickets"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/sprint-changelog-from-linear-claude-cowork"},{"slug":"sort-pdfs-into-folders-claude-cowork","title":"How to sort hundreds of PDFs into folders with Claude Cowork","category":"team-ops","platform":"Claude Cowork","tools":["Claude Cowork","A local folder of text-based PDFs","Excel (for the index)"],"problem":"A folder of hundreds of unsorted PDFs is useless until someone opens each file and files it.","steps":["Put all the PDFs in one local folder and select it in Cowork.","Write an explicit instruction: how many files there are, what text to look for, and which subfolder matching files go into.","Let Cowork read and move the files.","Optionally have it group related PDFs for merging.","Ask for an XLSX index showing where each file went."],"prompt_template":"This folder contains [NUMBER] PDF files. Sort them into subfolders:\n- [SUBFOLDER A]: files that mention [TEXT / CRITERIA]\n- [SUBFOLDER B]: files that mention [TEXT / CRITERIA]\n- _unsure: anything that doesn't clearly match\nDon't rename or delete files. When done, create index.xlsx with columns: original filename, destination folder, reason (the text you matched). Tell me how many went into each folder.","watch_outs":["Be very explicit in the prompt, as if briefing a smart new colleague.","Work on a copy of the folder the first time, and only turn it into a reusable skill if you'll repeat it."],"reported_result":"The user reports Cowork sorted 400 PDFs (382 in a detailed follow-up) without trouble. No measured accuracy or speed; it worked on text-based PDFs, not image-only ones.","source":{"url":"https://www.reddit.com/r/claude/comments/1ur3x8a/when_or_what_to_use_cowork_for","site":"Reddit","published":"2026-07-01"},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/sort-400-pdfs-into-folders","title":"Sort 400 PDFs into folders"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/sort-pdfs-into-folders-claude-cowork"},{"slug":"ai-github-pull-request-review-n8n","title":"How to review GitHub pull requests automatically with an n8n AI agent","category":"engineering-security","platform":"n8n","tools":["n8n","GitHub (trigger, API and nodes)","OpenAI chat model","Google Sheets (team best practices)"],"problem":"Small teams want a first review pass on every pull request, checked against their own coding conventions, before a human looks.","steps":["Add a GitHub Trigger node that fires when a PR is opened.","Call GitHub's pull-request files API with an HTTP Request node to get filenames and diffs.","Merge them into one prompt in a Code node, casting the model as a senior developer for your stack.","Give the agent a Google Sheet of team best practices as a tool.","Post the review as a PR comment and add a label (ReviewedByAI in the source)."],"prompt_template":"You are a senior [LANGUAGE/STACK] reviewer. Review this pull request diff against our team rules: {{best_practices}}.\nReport at most 5 findings, most important first. For each: file, approximate line, problem, why it matters, suggested fix. Skip style nitpicks a linter would catch. If the PR looks fine, say so in one sentence. Keep the whole review under 250 words.\n\nDiff: {{diff}}","watch_outs":["Validate each node's output as you build.","Inline comments need a loop over each file's diff and line numbers, which the author hadn't built yet."],"reported_result":"On a test PR the agent posted a comment and applied the label, but as one general comment rather than line-level feedback. The author found reviews often too verbose and sometimes missing the core problems.","source":{"url":"https://medium.com/@anirbanmukherjee1311/automate-github-code-reviews-with-n8ns-ai-agent-0b20be449c6c","site":"Medium","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/review-github-pull-requests-automatically","title":"Review GitHub pull requests automatically"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/ai-github-pull-request-review-n8n"},{"slug":"cve-monitoring-ai-scoring-alerts-n8n","title":"How to monitor new CVEs and get AI-scored alerts with n8n","category":"engineering-security","platform":"n8n","tools":["n8n","Bright Data (scraping)","Cohere Command-R","CISA KEV, CIRCL and AlienVault OTX (enrichment)","Slack","Gmail","Google Sheets"],"problem":"Security teams need to know quickly when a relevant vulnerability is published, without reading every advisory feed.","steps":["Scrape sources such as CISA, NIST, CERT-FR and BleepingComputer on a schedule.","Have an AI agent extract CVE details and score severity, enriched with CISA KEV, CIRCL and AlienVault OTX data.","Parse and normalise the output so scores stay consistent.","Remove duplicates with hash-based diff checks.","Send alerts to Slack and Gmail and log them in Google Sheets, with retries when a source fails."],"prompt_template":"Extract vulnerability details from the advisory below. Return JSON only:\n{\"cve_id\": \"...\", \"product\": \"...\", \"vendor\": \"...\", \"cvss\": number|null (only if stated), \"exploited_in_wild\": true|false|null, \"summary\": \"one sentence\", \"relevance\": \"high|medium|low\" for a stack using [YOUR STACK], \"reason\": \"...\"}\nDo not compute a CVSS score yourself; use null if none is given.\n\nAdvisory: {{text}}","watch_outs":["Without strict output constraints and parsing, the AI scored inconsistently.","Prefer official feeds and APIs over scraping where they exist."],"reported_result":"The author reports handling 100+ CVEs a day with alerts in under 5 minutes and 99.8% uptime; these are the author's own, unverified claims. Interactive Slack buttons were only a demo, so they call it about 90% production-ready.","source":{"url":"https://dev.to/joupify/soc-cert-automated-threat-intelligence-system-with-n8n-ai-5722","site":"DEV Community","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/monitor-and-score-cves-with-real-time-alerts","title":"Monitor and score CVEs with real-time alerts"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/cve-monitoring-ai-scoring-alerts-n8n"},{"slug":"kubernetes-alert-diagnosis-mcp-n8n","title":"How to diagnose Kubernetes alerts with an n8n agent and MCP servers","category":"engineering-security","platform":"n8n","tools":["n8n","Prometheus Alertmanager","Kubernetes, Grafana, DigitalOcean and GitHub MCP servers","Qdrant (internal docs)","Mattermost"],"problem":"When an alert fires at night, the on-call engineer spends the first minutes gathering context from several dashboards before they can start fixing anything.","steps":["Point an Alertmanager webhook at an n8n workflow.","Deduplicate alerts over a window (48 hours in the source) and fetch the triggering PromQL expression from the Prometheus Rules API.","Let an AI agent investigate read-only through Kubernetes, Grafana, cloud and GitHub MCP servers plus a vector store of internal docs.","Have it write a report: summary, event timeline, up to two root-cause hypotheses and troubleshooting steps.","Find the original alert post in Mattermost and reply in its thread."],"prompt_template":"An alert fired: {{alert_name}} with expression {{promql}} and labels {{labels}}.\nInvestigate using the read-only tools available (Kubernetes, Grafana, docs). Do not change anything.\nReply with: 1) a one-paragraph summary, 2) a timeline of relevant events, 3) up to two root-cause hypotheses with the evidence for each, 4) next troubleshooting steps for the on-call engineer. If a tool call fails, retry once, then continue without it and say what's missing.","watch_outs":["Limit tool retries to one per failed call to avoid loops, and strip noisy labels such as job and instance from the prompt.","Keep the agent read-only; you configure your own MCP endpoints and credentials."],"reported_result":"The source gives no measured results; it describes the report format. The author was still weighing extra data sources such as database stats and pipeline status.","source":{"url":"https://community.n8n.io/t/ai-alert-assistant-automated-incident-diagnostics-with-5-mcp-servers-alertmanager-ai-agent-mattermost/285369","site":"n8n Community forum","published":null},"via":{"name":"Agent Use Cases","url":"https://agentusecases.ai/use-case/diagnose-kubernetes-alerts-with-mcp-servers","title":"Diagnose Kubernetes alerts with MCP servers"},"added":"2026-10-15","last_checked":"2026-10-15","url":"https://recipes.branolabs.com/recipes/kubernetes-alert-diagnosis-mcp-n8n"}]}