LAP

LAP, short for Lean API Platform, supplies AI coding agents with compressed, agent oriented API specifications so they do not have to guess at endpoints from memory. It is built around a registry of pre compiled specs and a command line tool. A developer installs LAP into a coding tool, with setup commands for Claude Code, Cursor and Codex and a listing on ClawHub for OpenClaw. The agent can then fetch a verified spec covering endpoints, authentication and types for an API. The registry lists more than 1,500 APIs, including Stripe, GitHub, Slack, Discord, Twilio, OpenAI, Cloudflare and Anthropic. API providers can compile an OpenAPI, GraphQL or AsyncAPI spec into the LAP format, generate a skill that adds authentication and execution playbooks, and publish it to the registry. The site reports that specs are about ten times smaller than the originals and cites a benchmark of 500 runs across 50 APIs, with accuracy rising from 0.399 without specs to 0.860 with LAP. It is aimed at developers who use AI coding agents to integrate third party APIs, and at API providers who want agents to use their APIs correctly. The command line tool installs through npm or pip. The site publishes no pricing.