AI-assisted development reference

This section covers how to give AI coding assistants current Quake AI data while they work. The hosted MCP server exposes documentation search and infrastructure templates as callable tools, and a set of plain-text and JSON URLs serve the same documentation to tools that do not support MCP.

The MCP endpoint is https://docs.quake.ai/api/mcp. It requires no account and no API key, and it reads data only.

Pages in this section#

• What the MCP server is and why to use it: the conceptual overview. What the server is, why you would connect it, what search_docs, get_template, validate_launch_manifest, and prepare_launch each do, what "validated" means for a template, and a worked example.
• How to connect AI tools to Quake AI docs: setup for Cursor, Claude Desktop, Claude Code, VS Code (Copilot agent mode), Continue, and Zed, plus the plain-text documentation URLs for tools without MCP support.
• How to provision a server with an AI agent: drive a connected assistant to provision a running server. Prompt the agent, review the generated OpenTofu, and apply it yourself.
• How to save Quake AI knowledge to Obsidian, Git, or Notion: file MCP results into your notebook or repository, including the export_okf knowledge bundle.
• AI tool rules starter for Quake AI projects: an AGENTS.md plus per-tool adapter files that ground your assistant on Quake AI conventions.
• AI tools reference: the tool surface in detail: inputs, return shapes, no-match behavior, launch validation and handoff tools, the response envelope, JSON-RPC examples, and the knowledge-graph.json schema.
• Launch handoff: declare launch intent in quake.yaml, validate through MCP, and download a handoff packet for manual or control-plane execution.
• quake.yaml reference: manifest schema, field reference, and validation rules for launch handoff.

When to use this section#

Read these pages when you are:

• Setting up an AI coding assistant to write Quake AI infrastructure or application code and want it to work from current documentation rather than its training data.
• Deciding which retrieval surface fits your tool: the MCP server for MCP-capable assistants, or the plain-text URLs otherwise.
• Calling the MCP endpoint directly and need the tool inputs, the response envelope, or the JSON-RPC shape.
• Standardizing how your team's assistants refer to Quake AI across a project.