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

# MCP

> Connect your AI agent to Cala's MCP

```json theme={null}
https://api.cala.ai/mcp/
```

For more information about MCP and how it works, you can read more about it [here](https://modelcontextprotocol.io/docs/getting-started/intro).

<Note>
  **Pair with the agent skill.** The MCP server gives your agent access to Cala's tools. The [agent skill](/integrations/agent-skill) gives it the knowledge to use them well. Install both for the full experience.
</Note>

## Connect your agent to Cala's MCP

You can use Cala via our MCP. To get started, you need to get an API key to authenticate your MCP client. Go to our [Console](https://console.cala.ai/api-keys) and create a new API key.

<Card title="Get Cala's API key" icon="key" horizontal href="https://console.cala.ai/api-keys">
  Create a free Cala account and get your API key.
</Card>

With your API key in hand, you can connect your AI agent such as Cursor, Claude Code and others to our MCP server. Cala MCP gives AI agents access to verified, structured and typed knowledge — both as natural-language answers and structured data — as well as entity information.

<Tabs>
  <Tab title="Cursor">
    Add to `~/.cursor/mcp.json`:

    ```json theme={null}
    {
        "mcpServers": {
            "Cala": {
                "url": "https://api.cala.ai/mcp/",
                "headers": {
                    "X-API-KEY": "YOUR_CALA_API_KEY"
                }
            }
        }
    }
    ```

    For more information, follow their documentation [here](https://cursor.com/docs/context/mcp#installing-mcp-servers).
  </Tab>

  <Tab title="VS Code">
    Add to `.vscode/mcp.json`:

    ```json theme={null}
    {
        "servers": {
            "Cala": {
                "type": "http",
                "url": "https://api.cala.ai/mcp/",
                "headers": {
                    "X-API-KEY": "YOUR_CALA_API_KEY"
                }
            }
        }
    }
    ```

    For more information, follow their documentation [here](https://code.visualstudio.com/docs/copilot/customization/mcp-servers)
  </Tab>

  <Tab title="Claude Desktop">
    Claude Desktop uses OAuth, so you don't need an API key.

    In Claude Desktop, go to **Customize → Connectors → Add → Browse connectors**, search for Cala and click on **Connect to Claude**.

    <img src="https://mintcdn.com/calaai/WSudIyxF6Merq2Gf/images/claude-desktop-directory.png?fit=max&auto=format&n=WSudIyxF6Merq2Gf&q=85&s=455569d1960f6eb0156ed629edbf247c" alt="Claude Desktop Directory" width="1900" height="1096" data-path="images/claude-desktop-directory.png" />

    A **browser tab** will open for you to log in to your Cala account. After you **authorize**, return to Claude Desktop and open the Cala connector to confirm all tools are enabled. If not, enable them manually.
  </Tab>

  <Tab title="Claude Code">
    To install Cala as a Claude **connector**, run the following command in your terminal adding your API key:

    ```text theme={null}
    claude mcp add --transport http cala https://api.cala.ai/mcp/   --header "X-API-KEY: YOUR_CALA_API_KEY"
    ```

    You can also install Cala as a **plugin**, it is part of Claude [community marketplace](https://github.com/anthropics/claude-plugins-community/blob/main/.claude-plugin/marketplace.json) for plugins. Cala plugin for Claude Code includes our remote MCP server as well as the skill so that Claude Code can use the MCP tools optimally.

    To do so, run this in your Claude Code terminal to install the community marketplace:

    ```text theme={null}
    /plugin marketplace add anthropics/claude-plugins-community
    ```

    Then, install Cala plugin:

    ```text theme={null}
    /plugin install cala@claude-community
    ```

    Run this to activate the plugin:

    ```text theme={null}
    /reload-plugins
    ```
  </Tab>

  <Tab title="Mistral Vibe">
    Add to `~/.vibe/config.json`:

    ```json theme={null}
    {
      "mcpServers": {
        "Cala": {
          "url": "https://api.cala.ai/mcp",
          "headers": {
            "X-API-KEY": "YOUR_CALA_API_KEY"
          }
        }
      }
    }
    ```
  </Tab>

  <Tab title="Codex">
    Add to `~/.codex/config.toml`:

    ```toml theme={null}
    [mcp_servers.cala]
    enabled = true
    command = "npx"
    args = [
      "-y",
      "mcp-remote",
      "https://api.cala.ai/mcp",
      "--header",
      "X-API-Key:${CALA_API_KEY}"
    ]

    [mcp_servers.cala.env]
    CALA_API_KEY = "YOUR_CALA_API_KEY"
    ```
  </Tab>

  <Tab title="OpenAI">
    <Note>
      To support the full power of our graph's flexible schema, this MCP server utilizes dynamic JSON objects. This provides a "schema-less" experience for the LLM. Because of this, it is incompatible with OpenAI's "strict mode" so please ensure `strict: false` is set in your OpenAI specification.
    </Note>

    For more information, follow their documentation [here](https://developers.openai.com/api/docs/guides/tools-connectors-mcp)
  </Tab>

  <Tab title="Other">
    For clients with built-in remote MCP support:

    ```json theme={null}
    {
        "mcpServers": {
            "Cala": {
                "url": "https://api.cala.ai/mcp/",
                "headers": {
                    "X-API-KEY": "YOUR_CALA_API_KEY"
                }
            }
        }
    }
    ```

    For clients without built-in remote MCP support:

    ```json theme={null}
    {
        "mcpServers": {
            "Cala": {
                "command": "npx",
                "args": [
                    "mcp-remote",
                    "https://api.cala.ai/mcp/",
                    "--header",
                    "X-API-KEY: YOUR_CALA_API_KEY"
                ]
            }
        },
    }
    ```
  </Tab>
</Tabs>

## Available Tools

Here are the tools available to the Cala MCP:

<Card title="knowledge_search" icon="message-check" horizontal>
  Get a succinct, token-optimized answer in natural-language markdown, with sources, explainability, and matching entities. Accepts either a natural-language question or a Cala QL expression as input.
</Card>

<Card title="knowledge_query" icon="table" horizontal>
  Get verified knowledge as structured, typed JSON rows plus matching entities. Accepts either a Cala QL expression or a natural-language question as input.
</Card>

<Card title="entity_search" icon="grid-dividers" horizontal>
  Search entities by name with fuzzy matching.
</Card>

<Card title="entity_introspection" icon="microscope" horizontal>
  Get the field schema for an entity by its UUID. Returns the available properties, relationships, and numerical observations you can use when querying an entity.
</Card>

<Card title="entity_retrieval" icon="circle-info" horizontal>
  Retrieve information about an entity by its UUID.
</Card>

## Hubs & Marketplaces

You can also find Cala's MCP server listed on the following hubs and marketplaces:

<Card title="mcpservers.org" icon="globe" horizontal href="https://mcpservers.org/servers/docs-cala-ai-integrations-mcp">
  Browse Cala's MCP server listing on mcpservers.org.
</Card>

<Card title="mcp.so" icon="globe" horizontal href="https://mcp.so/server/cala">
  Browse Cala's MCP server listing on mcp.so.
</Card>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.