Turn any website into LLM-ready markdown in 200ms
Self-hosted Firecrawl alternative. Smart static → dynamic fallback, readability cleaning, and parallel crawling — without the per-token bill.
No signup · No API key for self-host · Playground is just a preview
git clone https://github.com/Michael-Obele/tomoshibi.git && cd tomoshibi
docker compose up -d 560 req/s p50 11ms vs Firecrawl self-hosted 1.9 req/s p50 5.4s — ~300× throughput. Same $0. See benchmark
Stop paying per scrape
Hosted APIs charge by the token and throttle you. Tomoshibi runs on your metal.
Rate limits, quotas, and surprise bills when you scale RAG.
Spawning a browser per scrape kills throughput.
Static scrapers miss React/Vue content.
Free playground vs self-hosted
The hosted playground is a preview — rate-limited, ephemeral, and shared. For anything real, run it yourself.
Try it in the browser — no install.
Unlimited, private, and fast — your infra, your data.
One command · Ports 7431–7435 · No clash with 3000/8080
From zero to markdown in 30 seconds
No SDK, no dashboard, no API key dance.
Clone & up
docker compose up -d
API 7431 · Web 7432 · MCP 7433 · Redis 7434 · SearXNG 7435
Scrape anything
POST /v1/scrape {url}
Smart mode auto-picks static or Chromedp
Get LLM-ready data
Markdown + metadata + links + optional screenshot
Crawl, batch, search, monitor — same API
Keep Tomoshibi free & open
Tomoshibi is MIT-licensed and built in the open. Sponsorship funds Chromedp upkeep, SearXNG tuning, and hobby-tier performance work — so self-hosting stays $0.
Stars drive discovery · Sponsors fund maintenance · Both keep the free tier alive
Self-host Tomoshibi today
One binary, one compose file, no vendor lock-in. Your data never leaves your machine.