Using Skywalk API with AI Agents

AI agents can read your AppFolio data and turn it into reports, dashboards, alerts, and integrations — Skywalk's simple JSON endpoints, single-header auth, and self-describing llms.txt catalog are designed to make that hand-off cheap. This page is the jumping-off point: pick the integration that matches what you're trying to do.

Pick your integration

Claude Chat Integration

Talk to your portfolio in Claude.ai. Connect the Skywalk MCP server in a few clicks — no code.

Claude Code Integration

Build a Skywalk-powered dashboard, ETL, or service. Claude Code writes the code; your app calls Skywalk at runtime.

Other AI Agent Integrations

Wiring Codex CLI, Cursor, Windsurf, ChatGPT, or another tool? Quick reference for the common ones.

Which one should I use?

If you want to…	Use	What it gives you
Browse and explore your data in plain English, see interactive charts and reports right in the chat	Claude Chat Integration	Skywalk shows up as a built-in tool in every Claude.ai conversation, via the MCP server
Build a real piece of software that uses Skywalk — a dashboard, a sync job, a webhook responder, an internal tool	Claude Code Integration	Claude Code scaffolds the project, writes the Skywalk client, and wires up async polling against llms.txt
Use a different AI tool — OpenAI Codex CLI, Cursor, Windsurf, ChatGPT, Cline, Zed, Aider, etc.	Other AI Agent Integrations	Per-tool setup steps for the common MCP-capable clients and llms.txt-based coding agents
Build a custom agent with the Claude API, OpenAI tool use, LangChain, or LlamaIndex	Manual Context	A pasteable API context block, ready-to-use tool definitions, and reference polling helpers in Python and TypeScript

Before you start

Every path needs the same two things:

1. A Skywalk API account with a connected AppFolio project — create one.
2. A Skywalk API key — generate one from your project with the scopes you need.

After that, follow the page above that matches your use case.

How the pieces fit together

Two patterns power every integration on this site:

• llms.txt — a single URL that describes the entire Skywalk API in plain text: base URL, auth header, every endpoint, response shape, and the async polling rule. Coding agents (Claude Code, Codex, Cursor, custom) fetch it and code against it directly. Because it's live, the agent always sees the current endpoint list.
• MCP server — a hosted server at https://mcp.skywalkapi.com that exposes the same data as native, typed tools for any Model Context Protocol client. Claude.ai uses this; other MCP clients can too.

Same underlying data and key — just two different ways to hand it to an agent depending on whether the agent is generating code or being the runtime.