Skip to main content

Other AI Agent Integrations

Most AI tools connect to Skywalk through one of two patterns:

  • MCP server at https://mcp.skywalkapi.com — native typed tools for any Model Context Protocol client. The first connection opens a Skywalk page where you paste your API key.
  • llms.txt — one URL describing every endpoint, the auth header, and the polling rule. Coding agents that can fetch URLs and run commands code directly against this.

MCP-capable clients​

Add https://mcp.skywalkapi.com as a remote MCP server in the client's settings and authenticate with your Skywalk API key. The tools then appear natively in that client.

  • Claude.ai — Settings → Connectors → Add connector. Full walkthrough on Claude Chat Integration.
  • ChatGPT — Connectors / GPT Builder (paid plans). Add as a custom MCP connector.
  • Cursor — Settings → MCP → Add MCP server (HTTP transport).
  • Windsurf — Cascade settings → MCP servers.
  • Cline (VS Code extension) — "Cline: Add MCP server" → Remote.
  • Zed — context_servers block in ~/.config/zed/settings.json.

For clients that only support stdio MCP servers, the mcp-remote npm package wraps a remote HTTP MCP server in a stdio process so the client can launch it.

Code-writing agents (use llms.txt)​

These work just like Claude Code Integration — point the agent at https://api.skywalkapi.com/llms.txt, export your key as SKYWALK_API_KEY, and let it write the code. The CLAUDE.md-style persistence trick and the security caveat about env vars on that page apply here too.

  • OpenAI Codex CLI — paste the llms.txt URL in your first prompt; Codex fetches it and runs curl to call endpoints.
  • GitHub Copilot Chat (agent mode) — reference llms.txt in your prompt; agent mode can fetch and execute commands.
  • Aider — drop the API contract into CONVENTIONS.md at the repo root so every Aider session inherits it.
  • Goose — define a recipe that fetches llms.txt and reads SKYWALK_API_KEY from the environment.

Building your own agent​

Wiring Skywalk into a custom agent built on the Claude API, OpenAI tool use, LangChain, or LlamaIndex? See Manual Context — it has a pasteable API context block, ready-to-use tool definitions in both Claude and OpenAI function-calling formats, and reference async-polling helpers in Python and TypeScript.

Don't see your tool?​

If your AI tool can do either of these, you're covered:

The Manual Context page also has a short pasteable summary that works as a fallback in any tool with a system prompt or instructions field.

MCP Server
Tool catalog and behavior of the hosted MCP server.
Manual Context
Pasteable context, tool definitions, polling helpers for custom builds.
AI Agents Overview
Decide which integration path fits your use case.