Self-hosted DataLens OSS MCP server
DataLens OSS API и MCP
A translator from the cloud DataLens API to internal open-source DataLens 2.9.0 calls, plus a Streamable HTTP MCP server on top
Medium risk
We rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.
Why this level
- The tokens grant read and write access to your self-hosted DataLens's workbooks and dashboards
- The containers join the existing docker network of your DataLens install
Install
Manual install
docker compose up -d --buildNeeds an already-running DataLens OSS and a filled-in .env; the MCP endpoint appears at http://127.0.0.1:8394/mcp.
This is third-party code. Review the repository files before installing.
What it does
The project solves a specific problem: the open-source version of DataLens (self-hosted, a docker compose with us, auth, control-api and meta-manager services) has no public API matching the cloud DataLens Public API format. The datalens-api service turns cloud-format POST /rpc/<method> calls into internal HTTP requests to the OSS services, and datalens-mcp on top of it exposes MCP tools over Streamable HTTP. The repository does not spin up the DataLens database, US or UI itself and does not touch the user's existing docker-compose.yaml; its containers just join the already-running docker network.
Who it is for. For teams running a self-hosted open-source DataLens who want to give an agent access to it in the same format as the cloud version.
Good fit when
- You already run DataLens OSS 2.9.0 via the official docker compose and need MCP on top of it
- You need access to workbooks, charts and dashboards of a self-hosted DataLens in the cloud API format
- You need a separate layer that does not touch your existing DataLens configuration
Not a fit when
- You use the cloud Yandex DataLens, not a self-hosted version: for that you need a server that talks to the cloud Public API directly
- You do not run DataLens OSS version 2.9.0: the README says plainly that other versions were not tested
Example request
List the workbooks in our self-hosted DataLens and open the most recently created dashboardLimitations
Tested only against DataLens OSS v2.9.0; the author does not promise other versions work. It needs an already-running DataLens via the official docker compose and deploys nothing itself. The default AUTH_* login and password in the OSS version are often admin/admin, and the README advises changing them. This is an unofficial project, unrelated to Yandex or datalens-tech/datalens.
How to disable. Stop the project's containers (docker compose down) and remove the server from your MCP client config.
MCP
- Transport
- http
- Authentication
- API key
| Environment variables | |
|---|---|
| US_MASTER_TOKEN required, secret | United Storage master token from your DataLens OSS .env. |
| AUTH_LOGIN required | Login for the DataLens OSS auth service, often admin by default. |
| AUTH_PASSWORD required, secret | Password for the auth service; the README advises changing it from the default. |
| MCP_AUTH_TOKEN required, secret | Token the MCP server checks in the Authorization header. |
Security check
- The tokens grant read and write access to your self-hosted DataLens's workbooks and dashboards
- The containers join the existing docker network of your DataLens install
README in short
The Russian README explains the project as a translator from the cloud API to open-source DataLens 2.9.0 calls, states its unofficial status and MIT license plainly, gives a table of the two services with default ports, a quick start with .env setup and a curl check, and links to separate docs on install, configuration, the RPC API, MCP, security and errors.
FAQ
Does the project spin up DataLens itself?
No, it needs an already-running DataLens OSS via the official docker compose; this repository only adds an API translator and MCP on top of it.
Does it work for cloud Yandex DataLens?
No, the project solves a problem specific to the open-source version; cloud DataLens already has its own Public API.
Related
166 skills for scientific work: bioinformatics, cheminformatics, clinical data, geospatial analysis and 100+ databases
Google's open-source MCP server for databases: ready tools for Postgres, MySQL, BigQuery, Spanner and more, plus custom tools in tools.yaml
Official Hugging Face skills: Hub operations via the hf CLI, datasets, model training, Spaces, evals and deployment
A visualization language for agents and an MCP server: neat charts from a simple semantic spec