Open Source

Stop rebuilding scrapers.
Turn any website into stable APIs.

Hermai is the open catalog of website APIs. Use it when your product needs the same site data again without keeping scrapers alive.

Try the playgroundBrowse 1971 schemasContribute a websiteTalk to us about a data source

Integrates with

OpenAI
Anthropic
Cursor
LangChain
Gemini
Kiro
CrewAI
Windsurf
Devin
LangGraph
OpenAI
Anthropic
Cursor
LangChain
Gemini
Kiro
CrewAI
Windsurf
Devin
LangGraph

Use cases

Your agent reads the sites that matter.

Government records, business filings, procurement, and live market data, delivered as structured, site-specific APIs that stay stable when the site changes.

govinfo.gov

Compliance agent uses GovInfo to verify subsidy claims against official records and public packages.

Sources
govinfo.gov
API CALLofficial records
Coverage: 4 endpoints

Check official package records, dates, and downloadable documents to support compliance review and fraud investigation workflows.

Read calls
searchlist_packages_in_collectionpackage_summarypackage_content

Benchmarks

Faster. Cleaner. Proven.

Real workflows on government and consumer sites where scrapers fail. Benchmarked against Firecrawl and WebFetch: cleaner data, structured fields, fewer failures.

Try in playground - no signup needed
USAspendingusaspending.gov

Search recent federal awards

Response time
Hermai AI
327ms

10 awards · ID, recipient, amount

Firecrawl
286ms

200 OK, but 0 awards from the JS app

WebFetch
blocked

JS app - empty payload, no awards

Agent tokens
Hermai AI
588

10 awards · ID, recipient, amount

Firecrawl
-

200 OK, but 0 awards from the JS app

WebFetch
-

JS app - empty payload, no awards

Gov rows June 2026 · Zillow/GitHub April 2026 · cold fetch, single connection

How it works

Install once. Ask anything. Get structured data.

No scrapers. No HTML parsing. No guessing. Your agent reads schemas, calls the site, and hands you clean data from the registry.

$
npx skills add hermai-ai/hermai-skills --skill hermai
01

Install the skill

Run the install command in any agent. Claude Code, Cursor, Windsurf, and 40+ tools.

✓ Found skill: hermai
✓ Installed to ~/.claude/skills/
✓ Covers calling + contributing
02

Ask anything

Your agent reads the schema, calls the real API, and handles auth. No scraping, no guessing.

"find 3-bedroom Airbnb listings
 in San Francisco for next weekend,
 group of 4"
03

Get structured data

Clean structured data, not HTML noise. Schemas discovered once, reused whenever you need the same site data.

{
  "listings": [
    { "title":   "Sunny Mission flat",
      "price":   412,
      "beds":    3,
      "rating":  4.91 }
  ],
  "total": 12
}

Contribute

Don't see your site? Add it for agents.

Hermai is an open catalog of website APIs. Anyone can add a schema with the CLI; every contribution makes the next agent more reliable. AGPL-3.0.

$
go install github.com/hermai-ai/hermai-cli/cmd/hermai@latest
Get the CLIContributor guide
# 1. Find the data paths behind the page
hermai detect https://example.com
hermai probe --body https://example.com/products/123 | hermai extract
 
# 2. Map public search and detail requests
hermai intercept https://example.com
 
# 3. Turn it into a schema you can reuse
hermai registry push schema.json

Using Claude Code, Cursor, or Codex? Add the hermai skill and let your agent run the CLI for you, with schema rules enforced. The contributor references load on demand.

$
npx skills add hermai-ai/hermai-skills --skill hermai

Roadmap

Where we're headed.

Request a feature
  • CLI + Schema catalog

    Search, pull, execute. Structured JSON back.

  • Hosted execution

    Run schemas through Hermai with an API key. No local setup required.

  • Community contributions

    Open submissions with verification and trust scoring

  • Health monitoring

    Continuous verification, auto-flagging stale schemas

  • Contributor profiles

    Public profiles for schema authors with contribution history

  • Community ratings & discussions

    Upvote, downvote, and discuss schemas with other agent builders

  • Site-hunter rankings

    Weekly leaderboard for top schema contributors

  • Schema composer

    Search and return available endpoints for the data you need

  • Usage history

    Review recent runs, usage, and handoff details from one place

  • Safe example mode

    Try a structured response before running a live request.

  • Guided error fixes

    Explain missing params, auth needs, and next steps in plain language.

  • Team visibility

    Saved runs, usage history, and clearer handoff from trial to production.

FAQ

Frequently asked questions

Short answers about the catalog, schemas, and how Hermai helps teams use stable website APIs.

What is Hermai?

Hermai is an open catalog of website APIs. Contributors add schemas for sites that do not offer a clean public API, so products and agents can use stable data instead of brittle scrapers.

How is this different from web scraping?

Scrapers depend on page HTML, so they break when a site changes its layout. Hermai schemas describe the API surface behind a site and return structured JSON.

Is Hermai free to use?

Yes. The schema catalog, CLI, and HTTP API are open source under AGPL-3.0. You can browse schemas and try the playground without signing up.

What sites does Hermai support?

The catalog covers commerce, media, travel, finance, government, developer tools, and more. Check the live catalog to see what is covered today.

Can I use Hermai for government and public-record data?

Yes. Government and public records are a flagship vertical: federal registers, business filings, procurement, and state and local open-data portals, delivered as structured, health-monitored APIs. See the Government solutions page for the compliance, KYB, and procurement workflows.

How do I add a new schema?

Install the CLI, inspect a site's API surface, then publish the schema to the registry. The hermai skill can guide your agent through the process.

Who maintains the schemas?

Schemas are contributed and maintained by the community. Hermai runs health checks and shows trust metadata so users can see which schemas are verified.