MCP server

Connect Claude Code, Claude Desktop, Cursor, Windsurf or VS Code to Datapoint and collect human feedback from inside an agent.

The Datapoint MCP server lets an agent run surveys, comparisons, ratings and rankings with real people, then read the results back into the conversation. It wraps the same API as the rest of these docs.

Connect your client

Claude Code

Install it as a plugin:

/plugin marketplace add impel-intelligence/datapoint-mcp
/plugin install datapoint@datapoint

To pick up new versions, run /plugin marketplace update datapoint and then /plugin update datapoint@datapoint.

Claude Desktop

Add the server to claude_desktop_config.json. On macOS the file is in ~/Library/Application Support/Claude/, on Windows in %APPDATA%\Claude\.

{
  "mcpServers": {
    "datapoint": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/impel-intelligence/datapoint-mcp.git",
        "datapoint-mcp"
      ]
    }
  }
}

Restart Claude Desktop after saving.

Cursor

Add the same mcpServers entry to ~/.cursor/mcp.json, or add it under Settings, MCP.

Windsurf

Add the same mcpServers entry to ~/.codeium/windsurf/mcp_config.json.

VS Code

Add the server to .vscode/mcp.json in your workspace. VS Code uses a servers key:

{
  "servers": {
    "datapoint": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/impel-intelligence/datapoint-mcp.git",
        "datapoint-mcp"
      ]
    }
  }
}

Any other MCP client

Run the server over stdio:

uvx --from git+https://github.com/impel-intelligence/datapoint-mcp.git datapoint-mcp

Authorize and verify

Ask the agent to run setup. It opens a browser window where you sign in to Datapoint. Then ask it to run check_balance. A working connection returns your account balance.

Give it a first task

Start with one small job so you can see the whole loop:

Upload logo_a.png and logo_b.png, then ask 20 people which logo they prefer.
Show me the plan and the cost before you launch anything.

The agent follows this order:

  1. upload_media for each local file. Raters cannot see local paths, so files are uploaded first and referenced as dp:// URIs. Public https:// URLs work without uploading.
  2. plan_survey with your description. It returns the task type, the number of items and the cost.
  3. create_survey once you confirm the plan. This reserves credits and starts collecting responses.
  4. check_survey for progress and aggregated results.

Tools

ToolWhat it does
setupSign in to your Datapoint account in the browser
upload_mediaUpload local images, audio or video and get dp:// references
plan_surveyDesign a survey from a plain-language description
create_surveyLaunch a survey from a plan
check_surveyStatus, progress and aggregated results
get_survey_responsesRaw responses per rater, paginated
list_surveysList your surveys
pause_surveyStop serving tasks for an active survey
resume_surveyResume a paused survey
cancel_surveyCancel a survey and refund unused reserved credits
check_balanceAccount balance
add_creditsOpen a checkout link to top up

If the tools do not appear

  • Check that uvx --version works in a terminal. If it does not, install uv and restart the client.
  • Restart the client after editing its config file. Most clients read MCP config only at startup.
  • Check the JSON for a trailing comma or a wrong top-level key. VS Code uses servers; the other clients use mcpServers.
  • If tools appear but calls fail with an authentication error, run setup again.
  • Quickstart for the same flow through the REST API.
  • Task types for what plan_survey can choose between.
  • Billing for how credits are reserved and refunded.