Skip to main content

​
The scenario

You’re debugging a 502 error on the payments webhook endpoint. Your team has a ChatGrid board with architecture docs, tutorials, and past incident postmortems. Here’s how you use it without leaving the terminal.

​
Step 1: Search for relevant knowledge

> Search the payments board for webhook 502 errors
Claude Code finds two relevant sources:
  • A postmortem from February about webhook timeouts
  • Jordan’s architecture notes about the retry system

​
Step 2: Get context on the architecture

> Based on our board, how is the webhook endpoint structured?
  What middleware does it go through?
Claude Code searches the board and responds with specifics from your team’s docs — not generic advice. 
> Search the board for Stripe webhook signature verification
Alex added a YouTube tutorial last week. The relevant section is returned automatically.

​
Step 4: Add what you learn

You figure out the issue: the webhook timeout was set to 5 seconds, but Stripe sometimes takes 8 seconds to send the full payload. Save that knowledge for the team: 
> Add this to the payments board:
  "Incident 2026-03-19: Webhook 502s caused by 5s timeout on
  /api/stripe/webhook. Stripe payloads for subscription events
  can take up to 8s. Fix: increased timeout to 15s in
  fly.toml [http_service] section. Also added retry-after
  header handling."
Now the next person who hits this issue will find your notes instantly.

​
Step 5: Ask a broader question

> Based on everything on our payments board, what are the top 3
  reliability risks for our webhook system?
AI synthesizes across all your board’s content — postmortems, architecture notes, tutorials — and gives you a prioritized answer.

​
The full loop

Search board  -->  Get context  -->  Fix the issue  -->  Add what you learned
                                                              |
                                                              v
                                                    Next person finds it
Every debugging session makes the board smarter. Knowledge compounds instead of disappearing.

​
More example prompts

What you’re doingWhat you say to Claude Code
Starting a new feature”What does our board say about the payments data model?”
Reviewing a PR”Search the board for our error handling conventions”
Debugging”Search the board for issues with Supabase connection pooling”
Making a decision”Based on our board, what are the pros/cons of Inngest vs Temporal?”
Onboarding”Summarize everything on the payments board for a new team member”
Need to set up MCP first? See the MCP Setup guide.