Releases: schlemperdev/shantycrawl-mcp
Release list
v1.0.0 — ShantyCrawl MCP
🚀 v1.0.0 — ShantyCrawl MCP
Firecrawl with a chill token footprint. A drop-in replacement for firecrawl-mcp that saves ~18,000 tokens per session by lazy-loading 22 advanced tools.
✨ Since v0.5.2
🎯 Implicit Tool Activation (The Big One)
Advanced tools now auto-activate on first use — call map directly without tool_enable and it works for that call, then auto-deactivates. Explicit tool_enable still persists until tool_disable. Best of both worlds: zero friction for one-off use, full control for heavy sessions.
🔧 Monitor Tools — v2 API Compliance
monitor_createbody format updated to Firecrawl v2 (targets: [{type:"scrape", urls: [...]}],schedule: {text|cron})monitor_updatenow has a dedicatedbuildBodymapper — no more raw-arg fallback
🧪 Full Test Suite
10 test files covering: smoke test, state persistence, activation lifecycle, prompts, error handling, scrape, search, crawl, advanced tools, and monitor CRUD. Zero external test frameworks — pure JSON-RPC over stdio.
🤖 CodeRabbit Automation
scripts/coderabbit-loop.tspolls PR issues for early green light (avoids waiting for full review)- Infinite loop bug in the polling logic fixed
- Enforced green-light rule before merge
🧹 And More
- Zero third-party wrappers — native
fetch, pure TypeScript FIRECRAWL_API_URLauto-detects local vs cloud fromFIRECRAWL_API_KEYpresence- All routes target Firecrawl v2 API
🔌 Quick Start
FIRECRAWL_API_URL=http://localhost:3002 npx shantycrawl-mcp⚙️ Claude Desktop
{
"mcpServers": {
"shantycrawl": {
"command": "npx",
"args": ["shantycrawl-mcp"],
"env": { "FIRECRAWL_API_URL": "http://localhost:3002" }
}
}
}📦 Install
npm install -g shantycrawl-mcp🙏 Special Thanks
To the Firecrawl team for the excellent API, and to everyone who beta-tested the 0.x releases.
Full changelog: v0.5.2...v1.0.0
v0.5.2 — compact JSON responses (token economy)
Replace JSON.stringify(d, null, 2) with JSON.stringify(d) in all 21 response handlers. Saves ~40% tokens in JSON responses with zero semantic change.
v0.5.1 — fix coderabbit-loop infinite loop + enforce green light rule
Changes
- coderabbit-loop.ts: add processed comment cache (prevents re-applying fixes), fallback green light detection
- AGENTS.md: add CodeRabbit Green Light Rule — no merge without exit 0
- skills/git-merge/SKILL.md: add Guard Clause 2 — verify green light before merge
v0.5.0 — tool_enable hint in core output
scrape/crawl/search output now appends hint: Need more? Call tool_enable() to list and activate advanced tools.
v0.4.0 — CodeRabbit Auto-Fix Loop
What's New
- scripts/coderabbit-loop.ts — Automated CodeRabbit fix loop for PRs
- Polls GitHub API for reviews, applies inline suggestions, commits/pushes
- Supports
suggestion,ts, and plain code fences - Descending line order per file prevents index shift
- MAX_LOOPS=5 safeguard with PR comment for manual intervention
- Detects CodeRabbit no-action signal when review body is edited in-place
- Uses execFileSync/argv (no shell injection) + paginated API calls
- AGENTS.md updated with coderabbit-loop work guidance entry
v0.3.2
Changes
Full migration to Firecrawl v2 API with buildBody architecture.
Fixes monitor_create sending v1-shaped body to v2 endpoint (closes #12).
Infrastructure
buildBodyargument shaping for all POST/PATCH tools — fallback-safe- Selective field copy prevents unknown args from reaching API
Path fixes (v1→v2)
- crawl, map, check_crawl_status, monitor_delete
Route/method fixes
- interact: /v2/browser → /v2/scrape
- interact_stop: POST → DELETE
- search_feedback: /v2/feedback → /v2/search/{id}/feedback
- All research tools: POST /v1/research/* → GET /v2/search/research/*
buildBody mappers
- scrape (formats strings→objects), crawl, search, map, extract, agent, parse, interact, monitor_create
Schema
- monitor_create: added optional schedule field
Diff
- 2 files changed, 140 insertions, 19 deletions
v0.3.1
- promote check_crawl_status to base tool
- tool_enable/tool_disable: no-arg lists available/active tools
- isBaseTool guard, invalid name suggestions
- keep prompts feature (shantycrawl-setup)
v0.3.0 – Cloud API Key Support
What's new
- Cloud API key support: set
FIRECRAWL_API_KEYenv var to authenticate with Firecrawl Cloud - Smart default URL: automatically selects
https://api.firecrawl.devwhen a key is detected,http://localhost:3002otherwise FIRECRAWL_API_URLalways overrides the default
Docs
- New README with comparison table, token savings (~18k per session), and known issues section
- AGENTS.md contracts updated for authentication and tool count
v0.2.0 — Tool discovery improvements
🚀 Enhancements
- MCP Prompts: New
shantycrawl-setupprompt with full usage guide for lazy-loading tools, accessible viaGetPromptRequestSchema. tool_enableDiscovery: Description now lists all advanced tool categories (map,extract,parse,agent,interact,research_*,monitor_*) so agents discover them naturally.- Call without args:
tool_enable()lists all 22 advanced tools with status;tool_disable()lists active tools.
🛡️ Guardrails
- Base tools locked:
scrape,crawl,search,check_crawl_status,tool_enable,tool_disablecannot be enabled/disabled. - Invalid name suggestions:
tool_enable("typo")returns available tool names.
🤖 Workflow & DX
- Git skills split: Separated
git-workflow(branch → PR) fromgit-merge(merge → version → publish) for cleaner automation. - SKILL.md shipped: OpenCode skill included in npm package — triggers on web scraping, mapping, extraction tasks.
📖 Documentation
- AGENTS.md guidance: Two lines in Work Guidance reminding agents to use
tool_enablefor advanced tools. - src/AGENTS.md: Updated file map and work guidance reflecting the new prompts capability.
v0.1.4
🐛 Bug Fixes & Agent Corrections
- npx Execution: Made the
dist/index.jsfile executable to ensure the binary works correctly across environments (chmod +x). - npm Compatibility: Removed the
./prefix from thebinpath inpackage.jsonto guarantee correct path resolution by npm.
🤖 Workflow & DX
- Safe Automation: Added the
git-workflowskill and updatedAGENTS.mdto structure the AI agent's workflow and protect themainbranch from direct commits.
🎨 Docs & Assets
- Branding: Final update of the official project avatar (ShantyCrawl mascot).
- Documentation: Minor formatting adjustments in
README.md.