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

# Condense MCP

> Connect the AI tool you already use to Condense and ask about your pipelines, connectors and logs in plain language. See what it can do, what you need and how access is controlled.

Condense MCP connects the AI tool you already use, such as Claude, Cursor or GitHub Copilot, to your Condense environment. You ask in plain language. The AI tool inspects your pipelines, reads connector configurations and logs, searches the Condense documentation for you, and can deploy a connector. It works as you, so it can only do what your Condense permissions already allow.

> [!NOTE]
>
> **Model Context Protocol (MCP)** is a standard way for an AI tool to discover and use the tools that another service offers. Condense MCP is the Condense side of that standard. You may also see an AI tool called an AI assistant or an MCP client.

## At a glance

| | |
| :- | :- |
| **Server URL** | `https://app.condense.io/api/mcp_servers/mcp` |
| **Tools** | 9 in total. You choose which 8 to allow when you connect. The ninth is internal and always on |
| **Connection** | A remote server over HTTP. There is nothing to install on your machine |
| **Sign-in** | OAuth, with your Condense account, in your browser |
| **Environment** | You pick one environment when you approve the connection |
| **Where you manage it** | Console, under **Condense MCP** |

## What you can do

> [card] See what is running
>
> Get the pipeline in a workspace: its connectors, the connections between them and their resource usage.

> [card] Read connector configurations
>
> See how a deployed connector is configured, or open a custom connector as an editable application.

> [card] Read connector logs
>
> Collect the logs of a deployed connector for a time window you choose.

> [card] Deploy connectors
>
> Find out which fields a prebuilt or custom connector needs, then deploy it with the configuration you give.

> [card] Ask the documentation
>
> Ask how something works in Condense and get the most relevant sections back.

> [card] Find your workspaces
>
> List the workspaces you can reach, or switch to the one you name.

## What it does not do

* **It cannot change much.** Deploying a connector is the only change it can make. It has no tools to restart, scale or delete anything, so do those in Condense Core or Console.
* **It does not read metrics, alerts or Kafka topics.** The nine tools cover workspaces, pipelines, connectors, connector logs and documentation. Use Condense Core and Console for metrics and alerts, and ask about the pipeline, its connectors and their logs instead of Kafka topics.
* **One connection reads one environment.** To use another environment, connect again and choose it.
* **It cannot give you more access than you have.** Your Condense permissions still apply to everything it does.

## Before you connect

You need:

* An active Condense account in the organization you want to use.
* Membership of the environment and workspace you want to work with, with the permissions you need there.
* An AI tool that supports remote MCP servers and the OAuth sign-in that Condense uses.
* A browser that can complete the sign-in and the approval.
* To deploy a connector, the connector's configuration values and a role that lets you deploy in that workspace.

## How access is decided

Three things decide what your AI tool can do, and the narrowest one wins.

```mermaid theme={null}
flowchart LR
    A["What you approve<br/>when you connect"] --> D(["What your AI tool can do"])
    B["What your administrator<br/>allows in the environment"] --> D
    C["What your Condense role<br/>already allows"] --> D
```

* **What you approve.** When you connect, you choose one environment and which tools the AI tool may use. You can change this later from the AI Tools page.
* **What your administrator allows.** An Organization Admin sets the allowed permissions and the operating mode for each environment in AI Governance.
* **What your Condense role allows.** The AI tool inherits your permissions. It can do nothing in Condense that you could not do yourself.

Access also follows the way Condense is organized.

| Level | What it means for Condense MCP |
| :- | :- |
| **Organization** | The organization you sign in to. It is shown on the approval page. |
| **Environment** | The one environment you choose on the approval page. The AI tool reads this environment. |
| **Workspace** | Every tool works inside a workspace. You name one, or the AI tool asks you to pick one. You only see workspaces you are authorized for. |

> [!TIP]
>
> Approving a connection never gives you anything new. If your AI tool cannot read something, the fix is a role from your administrator, not a change to a setting on your side.

## Next steps

> [card] [Quickstart](https://docs.condense.io/condense/v3.0.0/mcp/quickstart.md)
>
> Connect and ask your first question in a few minutes.

> [card] [Connect your AI tool](https://docs.condense.io/condense/v3.0.0/mcp/connect-your-ai-tool.md)
>
> Setup for Claude Code, Claude Desktop, Cursor, GitHub Copilot, Gemini CLI and Codex.

> [card] [Tool reference](https://docs.condense.io/condense/v3.0.0/mcp/tool-reference.md)
>
> The nine tools and what each one needs.


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