@indiekitai/trello-autopilot

Trello bug auto-fix CLI — scans bug cards from a Trello board, invokes a coding agent (e.g. Claude Code) to fix them, runs tests to verify, manages git branches/PRs, and moves fixed cards to "Done" with a summary comment.

Features (v0.2.0)

• Test verification — auto-runs npm test or pytest after fix; failed tests → card gets "fix-failed" label + failure details comment
• Git integration — creates fix/card-{id} branches, generates diff summaries, --pr mode for pull requests via gh CLI, git blame analysis
• Priority sorting — processes cards by label priority: critical > high > medium > low
• Smart filtering — --label critical to fix only specific labels, --limit N to cap count
• Failure handling — failed fixes get "needs-human" label + detailed comment with suggestions
• Retry — --retry to re-attempt previously failed cards (fix-failed / needs-human)
• Reporting — summary report with fix/fail/skip counts and timing; --json for structured output; --webhook to POST results

Install

npm install -g @indiekitai/trello-autopilot
# or run directly
npx @indiekitai/trello-autopilot --board "MyBoard" --list "Bugs" --repo ./myapp

Setup

Get your Trello API credentials at https://trello.com/power-ups/admin

export TRELLO_API_KEY="your-api-key"
export TRELLO_TOKEN="your-token"

CLI Usage

# Fix all bugs on "Cutie" board
trello-autopilot --board "Cutie" --repo /path/to/repo

# Fix only critical bugs, max 3
trello-autopilot --board "Cutie" --label critical --limit 3

# Create PRs instead of pushing to main
trello-autopilot --board "Cutie" --repo ./myapp --pr

# Retry previously failed cards
trello-autopilot --board "Cutie" --repo ./myapp --retry

# Custom test command
trello-autopilot --board "Cutie" --repo ./myapp --test-command "make test"

# JSON output + webhook notification
trello-autopilot --board "Cutie" --json --webhook https://hooks.slack.com/xxx

# Preview only (no changes)
trello-autopilot --board "Cutie" --dry-run

Options

| Flag | Short | Default | Description | |------|-------|---------|-------------| | --board | -b | (required) | Trello board name | | --list | -l | "Bugs" | Source list name | | --done | -d | "Done" | Destination list for fixed cards | | --repo | -r | cwd | Repository path | | --agent | -a | "claude" | Coding agent CLI command | | --dry-run | | false | Preview only | | --json | | false | JSON output | | --limit | -n | all | Max cards to process | | --label | | all | Only fix cards with this label | | --pr | | false | Create PR via gh CLI | | --retry | | false | Retry fix-failed/needs-human cards | | --test-command | -t | auto-detect | Custom test command | | --webhook | -w | | POST results to URL | | --help | -h | | Show help |

How It Works

1. Connects to Trello and finds the specified board/list
2. Sorts cards by priority labels (critical > high > medium > low)
3. Applies filters (--label, --limit, --retry)
4. For each card:
   • Creates a git branch fix/card-{id}
   • Runs git blame analysis for context
   • Invokes the coding agent with card details as prompt
   • Runs tests (npm test / pytest / custom command)
   • If tests pass: commits, pushes (or creates PR with --pr), moves card to Done
   • If tests fail: adds "fix-failed" label + failure comment, does NOT move card
   • If agent fails: adds "needs-human" label + detailed failure comment
5. Outputs summary report (or JSON with --json)
6. Sends webhook if --webhook is configured

MCP Server

Use as an MCP tool server for AI agents:

{
  "mcpServers": {
    "trello-autopilot": {
      "command": "npx",
      "args": ["@indiekitai/trello-autopilot/mcp"],
      "env": {
        "TRELLO_API_KEY": "your-key",
        "TRELLO_TOKEN": "your-token"
      }
    }
  }
}

MCP Tools

| Tool | Description | |------|-------------| | scan_bugs | Scan a Trello list for bug cards (with priority sorting & filtering) | | fix_bug | Fix a specific bug card with test verification and git integration | | move_card | Move a card to another list with optional comment | | retry_failed | Retry previously failed cards (fix-failed/needs-human) | | get_report | Run autopilot and return a structured report |

Programmatic API

import {
  TrelloClient,
  scanBugs,
  fixBug,
  sortByPriority,
  filterByLabel,
  generateReport,
} from "@indiekitai/trello-autopilot";

const client = new TrelloClient({
  apiKey: process.env.TRELLO_API_KEY!,
  token: process.env.TRELLO_TOKEN!,
});

let bugs = await scanBugs(client, "Cutie", "Bugs");
bugs = sortByPriority(bugs);
bugs = filterByLabel(bugs, "critical");
console.log(`Found ${bugs.length} critical bugs`);

Labels Used

| Label | Meaning | |-------|---------| | critical / high / medium / low | Priority for sorting | | fix-failed | Auto-fix was applied but tests failed | | needs-human | Auto-fix failed completely, needs manual intervention |

License

MIT