Pull Firebase Crashlytics data into your terminal and agentic coding workflows.
Crash data lives in the Firebase console. Your coding agent lives in the terminal. crashpull bridges the gap — list top crashes, inspect stack traces, and resolve issues without leaving your workflow. Output as human-readable tables or --format json for agents.
npx -y crashpull@latestRequires Node 18+. Zero runtime dependencies. Uses your existing Firebase CLI login — no extra auth setup.
# 1. Check prerequisites (firebase CLI, auth, API access)
npx -y crashpull@latest doctor
# 2. Link a Firebase project and Android app
npx -y crashpull@latest init
# 3. List top crashes (last 30 days)
npx -y crashpull@latest list
# 4. Inspect a specific crash with full stack trace
npx -y crashpull@latest show a1b2c3d4
# 5. Resolve an issue
npx -y crashpull@latest resolve a1b2c3d4
# 6. Get compact agent guide for LLM system prompts
npx -y crashpull@latest llmAll commands support --format json for machine-readable output:
# Top fatal crashes from the last 7 days
npx -y crashpull@latest list --type fatal --since 7d --format json
# Full stack trace as JSON
npx -y crashpull@latest show a1b2c3d4 --format json
# Pipe into your agent's context
npx -y crashpull@latest list --format json | claude -p "Which crash should we fix first?"
# Get compact agent guide for LLM system prompts
npx -y crashpull@latest llm | pbcopyAdd this to your project's CLAUDE.md to give your agent access to crash data:
# Crash data
- Run `npx -y crashpull@latest list --format json` to see current top crashes
- Run `npx -y crashpull@latest show <id> --format json` to get full stack trace
- Run `npx -y crashpull@latest resolve <id>` after fixing a crash| Command | Description |
|---|---|
doctor |
Preflight checks — firebase CLI, auth, API, config |
init |
Interactive setup — links Firebase project + app |
list |
Top issues with --type, --signal, --since, --limit filters |
show |
Issue detail + full stack trace |
resolve |
Mark an issue as resolved |
llm |
Compact agent guide — pipe into LLM system prompts |
help |
Usage info (also supports --format json) |
MIT