open source · mit · v0.1.0

The web toolkit for AI agents.

Search, scrape, crawl, and index the web — one static binary, zero API bills.

github ↗
15/15 js corpus
2.8× tighter md
996ms p50 scrape
mit license
works inside claude code · cursor · devin · windsurf · codex · any mcp client

js-corpus · go run ./bench · 2026-10-03

spec
scrape
page → clean markdown, 3-tier fetch, --fit token saver
3-tier
search
15-provider metasearch, weighted rrf, --scrape --highlights
rrf
index
sqlite fts5 + embeddings — your corpus, your box
fts5+vec
query
exa-style semantic search over your index
offline
crawl
goal-directed adaptive crawl, concurrency, robots-aware
robots
extract
css, schema.org, or llm via your endpoint (ollama ok)
css+llm
serve
http api, async jobs, webhooks, tenancy
api+jobs
mcp
17-tool mcp server, stdio + sse
17 tools

from url to answer, in-process.

Four hops, one binary — every stage writes to the same sqlite file it reads from.

pipeline — single process
url any site
scrape 3-tier fetch
md trafilatura
index fts5+vec
query bm25+cos
diff
app = FirecrawlApp(
− api_key="fc-…"
+ api_url="http://localhost:8080"
)

drop-in for firecrawl.

Repaint your base URL — the SDK keeps working.

routes
  • /v2/scrape
  • /v2/crawl
  • /v2/search
  • /v2/map
  • /v2/extract
  • /v2/agent
migration guide →
pricing

one binary, zero bills.

scraping api $83/mosearch api $7/1kbrowser infra $0.10/hr webx $0
community for the self-hoster
$0 forever
includes:
  • all 16 commands
  • 3 api keys
  • community support
  • self-hosted
go install
pro for team deployments
$49 /mo
everything in community, plus:
  • unlimited api keys
  • rbac
  • audit log
  • priority support
get pro
enterprise for scaled orgs
contact
everything in pro, plus:
  • per-deployment license
  • dedicated support
  • air-gap ok
talk to us

the free tier is a product, not a demo.