Skip to content
Docs · connect agent

Connect your coding agent.

Ordel's MCP server speaks the standard MCP OAuth flow, every client below connects with no API key in the config. The first tool call opens a browser tab, you sign in and click Approve, and the client stores the grant itself.

Prerequisites

  • Claude Code CLI installed and signed in (`npm i -g @anthropic-ai/claude-code`).
claude mcp add --transport http ordel /api/mcp

What happens on first use

  1. 1Adding the server does NOT trigger authentication by itself.
  2. 2Start Claude Code (`claude`), then run `/mcp` and select `ordel`.
  3. 3Choose Authenticate. Your browser opens Ordel's consent screen.
  4. 4Sign in (if needed) and click Approve.
  5. 5Back in the terminal, `ordel` shows as connected. The grant is stored and reused silently after that.

How to verify

  • ✓Run `/mcp` inside Claude Code: `ordel` shows as connected with its tool count.
  • ✓Ask it "list my Ordel projects" to confirm a live tool call.

After connecting: install the Ordel skill

Ask your agent to run the get_skill tool and save the result as ordel-skill.md in your workspace. It carries the full testing method - where to source requirements, how to ground generation in your app, and when your agent should drive the browser itself versus hand exploration to Ordel. Agents with the skill installed make noticeably better first passes; the connect-time instructions will tell your agent when a newer version is available.

Trouble connecting Claude Code? If Claude Code stops calling Ordel tools, run `/mcp` and reconnect `ordel`. This re-opens the browser approval. Being asked to re-approve after about 30 days of inactivity is expected, not an error.