Lap-Platform/LAP

Your agents are guessing at APIs. Give them the actual Agent-Native spec. 1500+ API's Ready To-Use skills, Compile any API spec into a lean, agent-native format. 10× smaller. OpenAPI, GraphQL, AsyncAPI, Protobuf, Postman.

What it solves

LLM agents often hallucinate API endpoints, parameters, and authentication flows because raw API documentation (like OpenAPI) is too verbose and structured for humans rather than models. This leads to low accuracy in API calls and high token costs due to the massive size of the documentation.

How it works

LAP acts as a compiler that transforms standard API specifications into a compressed, "agent-native" format. It uses a five-stage compression pipeline:

  1. Structural removal: Strips YAML scaffolding.
  2. Directive grammar: Replaces nested structures with flat, single-line declarations using @directives.
  3. Type compression: Shortens type declarations (e.g., type: string, format: uuid becomes str(uuid)).
  4. Redundancy elimination: Extracts shared fields via @common_fields and @type.
  5. Lean mode: Removes descriptions where LLMs can infer meaning from parameter names.

Who it’s for

Developers building AI agents and those using agent-integrated IDEs (like Claude Code, Cursor, and Codex) who want to improve their agent's API interaction accuracy while reducing latency and token costs.

Highlights

  • Significant Compression: Achieves a median 5.2x compression on OpenAPI specs, reducing costs by 35% and increasing speed by 29%.
  • Multi-Format Support: Compiles from OpenAPI, GraphQL, AsyncAPI, Protobuf, Postman, and Smithy.
  • Agent Skills: Includes a CLI (lapsh) to search a registry of 1,500+ pre-compiled specs and install them as skills directly into AI agents.
  • Zero Information Loss: Preserves all endpoints, parameters, and type constraints while providing typed contracts to prevent hallucinations.
  • Round-trip Capability: Can convert .lap files back into OpenAPI format.

Related

  • Project
    HKUDS/CLI-Anything

    CLI‑Anything is a Python framework that automatically generates agent‑friendly Click CLIs for any software, outputting JSON for easy parsing. It includes a 7‑phase generation pipeline (analyze → design → implement → test → document → publish) and a package manager called CLI‑Hub to browse, install, and run a growing catalog of pre‑built CLIs (e.g., GIMP, Blender, Zoom). The project targets AI agents (Claude Code, Cursor, OpenClaw, etc.) that need deterministic, composable command interfaces, and it ships with a comprehensive test suite and security hardenings.

  • Dispatch
    Why Humanising LLM Outputs Is Counterproductive for Agent Workflows

    Human‑focused output styles like Simplified Technical English cause lossy compression in LLM agents, hiding failures and reducing fidelity, so they should be applied only at the final human‑consumer boundary.

  • Project
    headroomlabs-ai/headroom

    A local context compression layer for AI agents that reduces prompt and response tokens for JSON, code, and text while preserving critical information.

  • Project
    virgiliojr94/book-to-skill

    A tool that converts technical books and documents into structured, on-demand AI agent skills, reducing token costs and hallucinations by avoiding full-context dumping.

  • Project
    dirac-run/dirac

    An open-source AI coding agent designed for complex software engineering, featuring precise hash-anchored editing and AST-based codebase manipulation to reduce token costs and improve reliability.