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

# How to connect an MCP client to Ask Legora

> Connect another AI client to Legora and receive answers from Agent.

Legora exposes a remote Model Context Protocol (MCP) server with the **Ask Legora** tool.

## Availability

Legora enables this connection for each organization. If you can't connect, contact Legora support to request access. Until access is confirmed, treat a failed connection as a permissions question rather than a setup error.

## Connect your MCP client

Use `https://mcp.legora.com/mcp` for production. Legora routes the connection to your
organization's processing region after you sign in, so you don't need to select a region.

For staging, use `https://mcp.staging.legora.com/mcp`.

When you connect, your client sends you to Legora to sign in and approve the access request. Most MCP clients that support OAuth connect without extra setup.

## Ask Legora

The tool accepts:

* `prompt`: The task for Agent.
* `project` (optional): A Legora project ID or a copied Legora project URL.
* `legal_research` (optional): One or two jurisdictions from the tool's jurisdiction list.

If you omit `project`, Agent uses your private workspace.

Use `legal_research` to pin Agent to legal research in specific jurisdictions. Agent automatically
chooses accessible internal collections for those jurisdictions. Jurisdiction-only selection does
not include virtual or dedicated-provider collections:

```json theme={null}
{
  "prompt": "Compare the limitation periods for this claim.",
  "legal_research": {
    "jurisdictions": ["Sweden", "EU"]
  }
}
```

If you omit `legal_research`, Agent decides which sources are relevant to the task.

After starting the Agent chat, Ask Legora waits up to 20 seconds for the answer. If it is ready, the
tool returns the answer and a link to the Legora chat.

A `running` result means Agent is still working or waiting for you to answer a question, approve an
action, or continue the task. Open the chat link to follow progress, review citations, or unblock
Agent. The MCP tool does not continue polling after it returns the link, and clients should not call
Ask Legora again for the same task or retry with a narrower prompt or legal-research scope.

A `failed` result is terminal. It returns `answer: null` and still includes the Legora chat link.
Open the link to inspect the chat and any available failure context.

Agent uses the same permissions and capabilities that you have inside Legora.