Build on the public discovery surface

Mengbi publishes a read-only discovery API for the public AI tools directory. It is intended for agents that need to search, browse, or cite tool information without opening a browser. The API validates locale, pagination, query length, and filter values. It returns stable JSON envelopes, request identifiers, and explicit error statuses so an integration can retry or stop safely.

Machine-readable files

• OpenAPI 3.1 specification — typed parameters, response schemas, operation IDs, and error responses for the public API.
• llms.txt — a concise index of the site, categories, tools, and agent resources.
• MCP manifest and Streamable HTTP MCP — read-only tools for searching, listing categories, and fetching one published tool.
• XML sitemap — canonical localized URLs for discovery and crawling.

API basics

Use https://mengbi-ai.net as the API server. The public API is available under/api, and the web origin also hosts the machine-readable files listed above. Start with GET /api/v1/search?q=writing&locale=enfor a focused query, or GET /api/v1/tools?locale=en&page=1&pageSize=24 for a paginated directory view. A tool detail link uses a stable English slug and includes active source citations. Do not assume that a missing result means a product does not exist; it may be unpublished, unverified, or unavailable in the requested locale.

Agent behavior and limits

Cache successful read responses briefly, respect 429 and Retry-After, keep page sizes small, and cite the official URL when presenting a recommendation. The directory is not a substitute for vendor terms, current pricing, safety review, or human approval. The public endpoints do not expose private account data, administration routes, crawler credentials, or unpublished moderation notes.

For corrections, accessibility issues, privacy requests, or source questions, see Contact Mengbi. For how the directory is maintained, see About Mengbi and the privacy page.