mcp-name: io.github.flexmeasures/flexmeasures

An MCP server for FlexMeasures, the open-source energy flexibility platform. It lets AI agents build and operate complete sites: create assets and sensors, post time-series data, trigger forecasting and scheduling jobs, poll job status, retrieve results, and manage report automations — all through the FlexMeasures REST API, so the server's authentication and permission model stays authoritative.

uvx flexmeasures-mcp          # run directly
# or
pipx install flexmeasures-mcp

Requires Python 3.10+ and a FlexMeasures instance (yours, or a hosted one).

* either email+password or an access token.

{
  "mcpServers": {
    "flexmeasures": {
      "command": "uvx",
      "args": ["flexmeasures-mcp"],
      "env": {
        "FLEXMEASURES_HOST": "company.flexmeasures.io",
        "FLEXMEASURES_EMAIL": "me@company.com",
        "FLEXMEASURES_PASSWORD": "..."
      }
    }
  }
}

claude mcp add flexmeasures \
  -e FLEXMEASURES_HOST=company.flexmeasures.io \
  -e FLEXMEASURES_EMAIL=me@company.com \
  -e FLEXMEASURES_PASSWORD=... \
  -- uvx flexmeasures-mcp

flexmeasures-mcp --transport streamable-http --host 0.0.0.0 --port 8100

⚠️ HTTP mode is single-identity: every connected client acts as the configured FlexMeasures user. Put an authenticating proxy in front before exposing it beyond localhost.

Meta: health_check, connection_info (+ gated authenticate). Assets: list_asset_types, list_assets, get_asset, create_asset, update_asset (+ gated delete_asset). Sensors & data: list_sensors, get_sensor, create_sensor, post_sensor_data, get_sensor_data. Jobs: trigger_forecast, get_forecast, trigger_schedule, get_schedule, get_job_status, list_asset_jobs. Automations: create_automation, create_report_automation, list_automations, get_automation, update_automation (+ gated delete_automation).

Tool names match the stable operationIds in the FlexMeasures OpenAPI spec (served by every instance at /ui/static/openapi-specs.json; the curated surface is marked with x-mcp-tool).

Async semantics: trigger_* tools return a job UUID immediately (the API responds 202); poll get_job_status until FINISHED, then fetch results with get_schedule / get_forecast. Older FlexMeasures servers (< v3.0-32 API) are also supported.

Use the bundled build_site_and_run_schedule prompt, or ask your agent:

Build a FlexMeasures site "Demo plant" with a 900 kWh / 500 kW battery, post tomorrow's day-ahead prices, and compute an optimized charging schedule.

A typical run: create_asset (site, then battery with parent_asset_id) → create_sensor (battery power in kW at PT15M; prices in EUR/MWh at PT1H) → update_asset to point the site's flex-context at the price sensor → post_sensor_data (24 hourly prices) → trigger_schedule with a storage flex-model (soc-at-start, soc-min/soc-max, power-capacity) → get_job_status until FINISHED → get_schedule → the agent summarizes IDs, data ranges, job UUIDs and the schedule.

Reports run only as cron automations in FlexMeasures — there is no run-now endpoint. create_report_automation sets one up (e.g. with PandasReporter); results appear on the report's output sensor(s) after the cron fires, readable via get_sensor_data. Requires a FlexMeasures server with automations support.

Third-party packages can add tools by declaring an entry point:

[project.entry-points."flexmeasures_mcp.tools"]
my_plugin = "my_pkg.mcp_tools:register"   # register(mcp: FastMCP, settings) -> None

Broken plugins are logged and skipped; they never prevent server startup.

uv venv && uv pip install -e ".[test]"
uv run pytest                 # unit tests (no server needed)
uv run pytest -m e2e          # end-to-end (see tests/e2e/conftest.py)

E2e tests need a live FlexMeasures stack with workers — use tests/e2e/docker-compose.yml (own Postgres + Redis), never a shared instance.

Apache 2.0