Docs / Getting started

Connect your agent

Dawnrose runs an MCP server that any client supporting Streamable HTTP can use. Your agent needs two things: the endpoint and a token.

URL     https://app.dawnrose.ai/mcp
Header  Authorization: Bearer YOUR_TOKEN

Create an agent token

1. In the dashboard, choose Connect agent in the sidebar.
2. Click Create agent token.
3. Copy the token. It is shown only once.

The setup snippets on that page fill in the token for you. Tokens start with drt_.

An agent token is limited to the workspace you had selected when you created it. Actions taken with it are marked as AI in the activity feed and highlighted on the canvas. To give an agent access to another workspace, switch to that workspace and create another token.

Set up your client

Replace YOUR_TOKEN with your agent token in each snippet.

Claude Code

Run in your terminal, then ask Claude to deploy something:

claude mcp add --transport http dawnrose \
    https://app.dawnrose.ai/mcp \
    --header "Authorization: Bearer YOUR_TOKEN"

Cursor

Add to ~/.cursor/mcp.json, or to .cursor/mcp.json in a project:

{
  "mcpServers": {
    "dawnrose": {
      "url": "https://app.dawnrose.ai/mcp",
      "headers": { "Authorization": "Bearer YOUR_TOKEN" }
    }
  }
}

Codex

Codex reads the token from an environment variable. Run in your terminal:

export DAWNROSE_TOKEN=YOUR_TOKEN
codex mcp add dawnrose --url https://app.dawnrose.ai/mcp \
    --bearer-token-env-var DAWNROSE_TOKEN

Codex reads DAWNROSE_TOKEN each time it starts, so put the export line in your shell profile (such as ~/.zshrc or ~/.bashrc) too.

VS Code (Copilot agent mode)

Add to .vscode/mcp.json, then use Copilot in agent mode:

{
  "servers": {
    "dawnrose": {
      "type": "http",
      "url": "https://app.dawnrose.ai/mcp",
      "headers": { "Authorization": "Bearer YOUR_TOKEN" }
    }
  }
}

Windsurf

Add to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "dawnrose": {
      "serverUrl": "https://app.dawnrose.ai/mcp",
      "headers": { "Authorization": "Bearer YOUR_TOKEN" }
    }
  }
}

Gemini CLI

Run in your terminal:

gemini mcp add --transport http dawnrose \
    https://app.dawnrose.ai/mcp \
    --header "Authorization: Bearer YOUR_TOKEN"

Claude Desktop

Claude Desktop connects through mcp-remote. Open Settings → Developer → Edit config, add this to claude_desktop_config.json and restart Claude Desktop:

{
  "mcpServers": {
    "dawnrose": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://app.dawnrose.ai/mcp",
               "--header", "Authorization: Bearer YOUR_TOKEN"]
    }
  }
}

opencode

Add to opencode.json:

{
  "mcp": {
    "dawnrose": {
      "type": "remote",
      "url": "https://app.dawnrose.ai/mcp",
      "headers": { "Authorization": "Bearer YOUR_TOKEN" }
    }
  }
}

Other clients

Any client that speaks MCP over Streamable HTTP works with the URL and header:

URL     https://app.dawnrose.ai/mcp
Header  Authorization: Bearer YOUR_TOKEN

Clients that only support stdio

The dawn CLI can act as a local stdio MCP server that forwards every message to Dawnrose:

dawn login --token YOUR_TOKEN --url https://app.dawnrose.ai
dawn mcp

dawn mcp config prints ready-to-paste configuration for Claude Code and for Claude Desktop using this bridge. If the server ends an idle MCP session, the bridge opens a new one by itself. See dawn CLI.

Check the connection

Ask your agent to call whoami. It returns your user, the workspaces the token can reach, your role and plan in each, and the available regions.

The Connect agent page lists every agent token under Your agents, with what the agent is doing right now or when it was last active, and shows the recent agent activity below.

The Connect page also lists example requests to try, such as Why is api crashing? or Create a staging environment from production and deploy it.

What the agent gets

Besides the tools, the MCP server sends the agent instructions on how Dawnrose works and offers two prompts: Deploy an app and Debug a failing service. It also exposes short reference documents on getting started, variables, dawnrose.json and the log filter syntax.

Errors returned to the agent carry a hint with the next step when there is one. Tools that delete or overwrite data require an explicit confirm: true argument. See Security and access.

Personal API tokens

Tokens created under Account → Tokens are meant for the CLI, CI pipelines and scripts. They are not limited to one workspace. If you use one over MCP, Dawnrose still treats the caller as an agent and names it after you (for example, “Ada’s agent”), so its work is marked as AI in the dashboard.

An agent token cannot be used to create new tokens.

Revoke a token

On the Connect agent page, click the trash icon next to the token under Your agents and confirm Revoke. Agents using the token lose access immediately. Account → Tokens lists all your tokens, including agent tokens, and you can revoke any of them there.

Running dawn logout removes the token from your machine but does not revoke it. Revoke it in the dashboard if it leaked.