MarkApiDown integrates directly with the AI coding agents you already use. RunningDocumentation Index
Fetch the complete documentation index at: https://docs.markapidown.net/llms.txt
Use this file to discover all available pages before exploring further.
mad install skills detects which agents are configured in your project and installs the mad skill for each one — no manual copying, no config editing. From that point on, your agent understands the MarkApiDown spec format, project layout, and workflow rules without any prompting from you.
Supported agents
MarkApiDown detects and installs skills for the following agents. Detection is based on the presence of the agent’s config directory in your project root or home directory.| Agent | Config directory | Detection path |
|---|---|---|
| Claude Code | .claude/ | <project>/.claude/ |
| Codex CLI | .codex/ | <project>/.codex/ or ~/.codex/ |
| Cursor | .cursor/ | <project>/.cursor/ |
| GitHub Copilot | .github/ | <project>/.github/ |
| Antigravity | .agents/ | <project>/.agents/, <project>/.agent/, or ~/.gemini/antigravity/ |
| OpenCode | .opencode/ | <project>/.opencode/ |
Skills are plain files written into the agent’s config directory. No background processes run and no network calls are made during installation. You can inspect or commit the installed files like any other project file.
Install skills
Install for all detected agents
Run the following command from your project root. MarkApiDown scans for all supported agent config directories and writes the If no agent config directories are found, you’ll see a clear error message with a fix suggestion. Either create the agent’s config directory first or pass
mad skill into each one it finds.--agent explicitly.Install for one specific agent
Target a single agent by name when you want to add or refresh a skill for just that tool.Writes to
- Claude Code
- Codex CLI
- Cursor
- GitHub Copilot
- Antigravity
- OpenCode
.claude/skills/mad/SKILL.md.Install slash commands (Claude Code and Codex CLI only)
Slash commands give Claude Code and Codex CLI the For a specific agent:Other agents use skills and MCP tools instead of slash commands. See Slash Commands for a full reference.
/mad and /mad-debug shortcuts for scanning routes, enriching specs, building flows, and diagnosing failures.Register the MCP server (Claude Code)
The MCP server exposes structured JSON tools that let Claude Code list, read, validate, execute, and author specs without parsing shell output. Register it with one command:This runs See MCP Tools for a full tool reference.
claude mcp add mad -- mad mcp under the hood. Verify the registration with:Check installation status
Run the following command to see which agents MarkApiDown detects in your project and whether themad skill is currently installed for each one.
Keep skills up to date
After upgrading MarkApiDown, refresh all installed skills and verify your project setup in one step:mad doctor checks whether installed skill files match the current version shipped with the CLI. The --fix flag overwrites any out-of-date files automatically.
What happens after installation
Once skills are installed, your agent automatically applies the MarkApiDown workflow whenever it touches files inapi-docs/. You don’t need to explain the spec format, frontmatter fields, or variable syntax — the skill covers all of that. To verify, open your agent and ask it to document an endpoint: it should produce a correctly structured spec without any additional prompting.
Skills reference
Learn exactly what the mad skill teaches your agent about spec format, project layout, and tool usage.
MCP tools
Explore the structured JSON tools the MCP server exposes to Claude Code.
Slash commands
Use /mad and /mad-debug in Claude Code and Codex CLI for one-shot workflows.
CLI Reference
Diagnose and fix project setup issues using mad doctor —fix.