> ## 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.

# Ask Condense MCP

> How to phrase requests, how your AI tool picks a workspace, and example prompts for pipelines, connectors, logs and documentation.

You do not call Condense MCP tools yourself. You ask your AI tool in plain language, and it picks the tools it needs.

## Name the workspace

Every tool works in a workspace.

* If you name a workspace in your request, your AI tool uses it.
* If you do not, it reuses the workspace from earlier in the conversation, or asks you to pick one.

Your AI tool finds your workspaces with the `get_all_workspaces` and `ensure_workspace` tools. You never have to look up a workspace ID first.

## How one question becomes several tool calls

A single question often uses more than one tool. For example:

```mermaid theme={null}
sequenceDiagram
    actor You
    participant Tool as Your AI tool
    participant MCP as Condense MCP
    You->>Tool: Why is the fleet-tracking pipeline quiet?
    Tool->>MCP: ensure_workspace
    Tool->>MCP: get_current_pipeline
    Tool->>MCP: get_connector_logs
    MCP-->>Tool: Pipeline, connectors and logs
    Tool-->>You: A summary of what it found
```

You may see several calls go by before the answer arrives. That is your AI tool working through the question.

## What to ask

These requests work today. Replace the names with your own.

### See what is running

* "Show me the current pipeline in the fleet-tracking workspace." Tool: `get_current_pipeline`.
* "Which workspaces do I have access to?" Tool: `get_all_workspaces`.
* "Switch to the fleet-tracking workspace." Tool: `ensure_workspace`.

### Check a connector's configuration

* "Show the config of connector test-connector in current pipeline." Tool: `get_connector_config_deployed`.

### Read connector logs

* "Get the last 5000 ms of logs for config ID 872 in the fleet-tracking workspace." Tool: `get_connector_logs`.

### Deploy a connector

* "What do I need to configure to deploy the NATS to Apache Kafka output connector?" Tool: `get_connector_configs_to_deploy`.
* "What config fields does my custom input connector Saas-Application need?" Tool: `get_custom_connector_configs`.
* "Deploy an NATS to MongoDB connector in fleet-tracking." Tool: `deploy_connector`.

### Ask how something works

* "How does RBAC work in Condense?" Tool: `query_docs`.

## Tips for better answers

* **Name the workspace.** It saves a round of questions.
* **Give a log duration.** Connector logs are collected for a duration in milliseconds, and the default is 200. Ask for what you need, for example "the last 5000 ms".
* **Say what kind of connector it is.** To find the fields a prebuilt connector needs, give its name and its type: Input, Output, Transform or Geofence Transform.
* **Give the connector's ID and category.** To read the configuration of a deployed connector, your AI tool needs its ID and its category.
* **Find out what a connector needs before you deploy it.** Ask for the configuration fields first, then deploy with those values.
* **Check before you deploy.** `deploy_connector` changes your environment. Read the configuration your AI tool is about to send before you approve it.

> [!WARNING]
>
> Connector configurations can include credentials such as broker addresses, usernames and passwords. Keep real secrets out of the chat where you can.

## Related

> [card] [Tool reference](https://docs.condense.io/condense/v3.0.0/mcp/tool-reference.md)
>
> What each tool does, and which one changes your environment.

> [card] [Scope and permissions](https://docs.condense.io/condense/v3.0.0/mcp/scope-and-permissions.md)
>
> Why your AI tool can or cannot do something.


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