> ## Documentation Index
> Fetch the complete documentation index at: https://docs.clarky.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# MCP Server

> Connect Claude, Codex, and other AI tools directly to your Clarky CRM

Clarky's MCP server lets AI tools you already use, like Claude and Codex, work with your CRM directly. Ask Claude about your pipeline, look up contacts, check analytics, or prepare an email draft without leaving your AI tool. It is the same brain as [Ask Clarky](/features/crm/ask-clarky), available everywhere MCP is supported.

Clarky implements the newest MCP specification (the 2026-07-28 revision) and remains backwards compatible with clients on earlier versions of the protocol.

<Info>
  **Plan availability:** Standard or higher (included with API Access). See [Plans & Billing](/features/settings/plans).
</Info>

## What your AI tools can do

Once connected, your AI tool has live access to this workspace's CRM:

* **Contacts and companies**: search, filter by tags or lead source, and pull full contact details
* **Pipeline**: stage summaries, deal values, and moving cards between stages
* **Analytics**: contact counts, deal totals, conversations, bookings, and form submissions over any date range
* **Bookings and meetings**: upcoming and past meetings
* **Conversations and activities**: chat, email, text, and voice history plus logged notes, calls, and tasks
* **Drafts**: prepare email, text, and sequence drafts to review and send inside Clarky

<Note>
  AI tools can never send emails or texts through the MCP connection. Drafts always come back to Clarky, where you review and click Send yourself.
</Note>

## Connect Claude (claude.ai and Claude Desktop)

<Steps>
  <Step title="Add a custom connector">
    In Claude, open Settings, then Connectors, and choose Add custom connector. Paste your server URL: `https://clarky.ai/api/mcp`
  </Step>

  <Step title="Sign in to Clarky">
    Claude opens a Clarky sign-in page. Log in with your normal Clarky account and pick the workspace you want to connect.
  </Step>

  <Step title="Approve access">
    Review the requested access and click Allow. Claude can now use your CRM in any conversation.
  </Step>
</Steps>

## Connect Claude Code

Run this in your terminal:

```bash theme={null}
claude mcp add --transport http clarky https://clarky.ai/api/mcp
```

Then run `/mcp` inside Claude Code to sign in to Clarky.

## Connect Codex

```bash theme={null}
codex mcp add clarky https://clarky.ai/api/mcp
```

## Connect with an API key instead

For headless or shared setups where a sign-in flow is inconvenient, use a workspace API key as the bearer token. Create one in Settings under MCP Server, then:

```bash theme={null}
claude mcp add --transport http clarky https://clarky.ai/api/mcp --header "Authorization: Bearer ck_live_..."
```

<Tip>
  Read-only keys are a good fit for shared or automated setups: the AI can look things up but cannot change pipeline stages or log activities.
</Tip>

## Managing connections

The MCP Server section in Settings shows every AI tool connected to your workspace, who connected it, and when it was last used. Disconnect any of them at any time; the tool will need to sign in again to regain access. API keys are managed separately in the API Keys section.

Access also ends automatically when a user is removed from the workspace.

## Tips for better answers

* Ask your AI tool to call `get-crm-context` first when a question mentions tags, pipeline stages, or lead sources by an approximate name. It returns the exact values in your workspace.
* If your workspace has more than one agent, mention which agent's CRM you mean, or ask for `list-agents` to see the options.