# Anthropic Claude

## What you need

- a Claude environment that supports remote MCP servers
- your Colabra account
- the production remote MCP URL:

```text
https://api.colabra.ai/mcp
```

![Anthropic Claude Cowork MCP](/docs/screenshots/claude.png)

## When to use Claude with Colabra

Claude is a strong fit when the team wants longer-form reasoning or writing against live project context: issue summaries, workstream syntheses, market-context overlays, and report drafting that still needs to stay grounded in the evidence base.

  External research plus internal evidence
  Real deal example: market context without losing the deal record

  A team wants to compare target claims against external market context, then roll that into an evidence-backed draft memo. Claude can do the synthesis, while Colabra MCP keeps the actual files, findings, requests, and report draft anchored to the live project.

## Setup in Claude Code

1. Add the Colabra server as a remote HTTP MCP server:

```bash
claude mcp add --transport http colabra https://api.colabra.ai/mcp
```

2. Start Claude Code and run `/mcp`.
3. Select the Colabra server and follow the browser OAuth flow.
4. After authentication, Claude Code can call the Colabra tools directly from the project session.

## Setup in Claude or Claude Cowork

1. Add a new remote MCP server in the Claude product you are using.
2. Select **remote** or **HTTP MCP** if Claude asks for a transport type.
3. Paste `https://api.colabra.ai/mcp`.
4. If Claude asks for manual OAuth fields, use the endpoints on the [MCP](/docs/mcp-and-integrations/mcp/) page.
5. Sign in with Colabra and approve access.

## Result

Claude can then use Colabra through MCP against the user-approved workspace context.

Ask it to inspect files, findings, requests, and reports directly rather than pasting excerpts into chat. That is where the connection stops being a convenience and starts becoming a real workflow improvement.