Connecting Claude Code to your Turtini workspace (the /turtini plugin)

Turtini ships an official Claude Code plugin that wires the @turtini/mcp server into your IDE in five seconds and adds a /turtini slash command for guided setup, key minting, verification, and troubleshooting.

This is the recommended path for Claude Code users — it removes the need to hand-edit any MCP config. (For Cursor, Windsurf, and Continue, the manual setup in "@turtini/mcp — letting AI-IDE agents collaborate with your Turtini org" is still the right path.)

Marketing + docs page: https://turtini.com/integrations/claude-code

Five-second install:
claude plugin install github.com/Turtini/claude-code-plugin

What that does:
• Registers a Claude Code plugin named "turtini" in your local plugin store
• Adds an MCP server entry that runs npx -y @turtini/mcp on demand
• Ships a /turtini slash command (this is the plugin's skill) into every Claude Code session

The plugin reads your Turtini API key from the TURTINI_API_KEY environment variable on your machine. The key is never bundled into the plugin and never sent to Anthropic — only to api.turtini.com when the agent actively calls a tool.

After install — three things to do once:
1. Mint a Turtini API key
In the Turtini app: Settings → Partner Dev → API Keys → Create key. Pick the scopes you want the agent to have (start with reads — contacts:read, accounts:read, opportunities:read — and add writes when you want the agent to actually ship). The plaintext key (turtini_…) is shown ONCE on creation. Copy it before you close the dialog.
2. Set TURTINI_API_KEY in your shell
zsh / bash: echo 'export TURTINI_API_KEY=turtini_xxx' >> ~/.zshrc && source ~/.zshrc
fish: echo 'set -gx TURTINI_API_KEY turtini_xxx' >> ~/.config/fish/config.fish
PowerShell: [System.Environment]::SetEnvironmentVariable('TURTINI_API_KEY','turtini_xxx','User')
3. Restart Claude Code so the MCP subprocess picks up the new env var, then verify with: claude mcp list — turtini should show "connected".

The /turtini slash command:
Type /turtini in any Claude Code session and the skill walks you through the same flow above based on what you say. It also handles:
• "rotate my key" — directs you to mint a new key + revoke the old one
• "verify the connection" — runs claude mcp list, diagnoses pending / failed states, hints at the env-var fix
• "what tools does it have" — lists the live MCP tool surface so you know what to ask for
• "uninstall" — gives you the right commands to disable temporarily or remove entirely

What the agent can do once connected:
The MCP tool surface is the same as the @turtini/mcp server — auth_whoami, contact_list / contact_create, account_list, opportunity_list, article_list / article_create, event_list / event_create. The full per-tool details are in the @turtini/mcp article and on the Public API docs page (turtini.com/docs/api).

Updates: the toolset grows automatically as @turtini/mcp publishes new versions — Claude Code fetches the latest @turtini/mcp every time it starts, so you don't re-install the plugin to get new tools.

Security model:
• Your API key lives in your shell env, nowhere else.
• Scopes on the key bound everything the agent can do — a read-only key cannot accidentally create records.
• Revocation is immediate: delete the key in Settings → Partner Dev → API Keys (or claude plugin uninstall turtini), and the agent loses access on its next call.

Source + issues: https://github.com/Turtini/claude-code-plugin/issues
Anthropic plugin marketplace listing: pending review — once approved, you'll be able to install with /plugin install turtini directly inside Claude Code.

See also:
• "@turtini/mcp — letting AI-IDE agents collaborate with your Turtini org" — manual install for Cursor / Windsurf / Continue
• "Generating an API key for the public REST API" — scope reference and key management
• "Org-context queries — natural-language Q&A over your Turtini org" — the org_context_query tool the agent picks up automatically