fireclaude

Self-hosted Firecrawl for Claude Code — no API keys, no paid tiers.

Self-hosted Firecrawl packaged as a Claude Code skill. Scrape, crawl, search, batch-dataset, JSON-schema extract, change tracking, webhook testing — all local, no API keys, no paid tiers.

Stack: Firecrawl API + Playwright (stealth) + SearxNG (search) + Ollama (llama3.2:3b, JSON extract).

Install

npm install -g fireclaude
fireclaude setup

fireclaude setup clones firecrawl, applies patches, builds Docker images, pulls the Ollama model, and runs smoke tests. Takes 5–15 min on first run. The running stack uses ~4 GB RAM and ~8 GB disk.

Override install location or model:

fireclaude setup --install-dir /opt/firecrawl --model qwen2.5:7b

Optional short alias: if you prefer typing fc, run:

fireclaude alias install

This writes alias fc='fireclaude' to ~/.bashrc / ~/.zshrc. Remove with fireclaude alias uninstall.

Claude Code plugin install

/plugin marketplace add sannidhyas/fireclaude

Or from a local checkout:

/plugin add ./fireclaude

The skill activates automatically when you ask Claude to scrape, crawl, search the web, or extract structured data.

CLI reference

Lifecycle

Data

Agentic use

Agent-safe JSON endpoints:

# Dep + health check — returns {deps, containers, models}
fireclaude doctor --json | jq '.deps.docker'

# Container status — returns [{service, state}, ...]
fireclaude status --json | jq 'length'

# Scrape
fireclaude scrape https://example.com --json | jq '.data.markdown'

# Search
fireclaude search "rust async runtime" --limit 5 --json | jq '.data[].url'

# Map a site
fireclaude map https://docs.anthropic.com --json | jq '.links | length'

# Extract structured data
fireclaude extract https://news.ycombinator.com \
  --schema schema.json \
  --prompt "Extract the top 3 story titles and URLs" \
  --json | jq '.data.json'

# Change tracking
fireclaude changes https://example.com --json
# → {"url":"...","changed":false,"current_hash":"...","diff_bytes":0,"scraped_at":...}

JSON-schema extract

cat > schema.json <<'EOF'
{
  "type": "object",
  "properties": {
    "top_story_title": {"type": "string"},
    "top_story_url":   {"type": "string"},
    "point_count":     {"type": "integer"}
  },
  "required": ["top_story_title"]
}
EOF

fireclaude extract https://news.ycombinator.com \
  --schema schema.json \
  --prompt "Extract the top story." \
  --json | jq '.data.json'

Routes through the ollama container (llama3.2:3b by default). First request cold-starts the model (~5–10s).

Model swap

fireclaude model list
fireclaude model pull qwen2.5:7b
fireclaude model swap qwen2.5:7b
fireclaude model current

Change tracking

fireclaude changes https://example.com          # first run → changed
fireclaude changes https://example.com          # second run → unchanged
fireclaude changes https://example.com --diff   # show unified diff
fireclaude changes https://example.com --json   # agent-friendly output

Change history stored in SQLite at ~/.fireclaude/changes.db (override with CHANGES_DB_PATH).

Webhook testing

fireclaude webhook-listen --json &
# Each POST to :4321/webhook emits a JSON line:
# {"timestamp":"...","method":"POST","path":"/webhook","headers":{...},"body":{...}}

fireclaude webhook-listen --port 9999

Turnstile / CAPTCHA solver (optional)

Add to your .env (disabled by default):

TURNSTILE_SOLVER=capmonster       # or: 2captcha
TURNSTILE_SOLVER_API_KEY=your_key

Honest limits

How patches work

Six git diff patches in docker/patches/ are applied against firecrawl SHA 0ae6387b762c7450190eb7d8f9f7b81b7adfcaab:

Uninstall

fireclaude teardown           # stop + prompt to remove clone + volumes
fireclaude teardown --purge   # stop + remove everything without prompts

Versioning

fireclaude follows SemVer. The project is currently pre-stable (0.x). See VERSIONING.md for the 0.x → 1.0.0 stability checklist.

Contributing

See CONTRIBUTING.md.

Security

See SECURITY.md.

License

MIT — see LICENSE.

# pinned release
curl -sSL https://raw.githubusercontent.com/sannidhyas/fireclaude/v0.5.0/install.sh | bash

# or clone first
git clone https://github.com/sannidhyas/fireclaude
cd fireclaude && ./install.sh

FIRECRAWL_INSTALL_DIR=/opt/firecrawl ./install.sh