Baseten MCP server

What the model can do

Baseten model serving and training: models with their deployments, environments and replica counts, deployment metrics, training projects and jobs, and the workspace audit log; can activate, deactivate, retry, rescale, promote or delete deployments, cancel a promotion, rename a model, and stop or recreate training jobs.

Reads

• baseten_list_modelsBaseten models: id, name, how many deployments, the production and development deployment ids, instance type and team.
• baseten_list_deploymentsDeployments of one Baseten model: status (ACTIVE, SCALED_TO_ZERO, INACTIVE, DEPLOY_FAILED…), whether it is production or development…
• baseten_list_environmentsEnvironments of one Baseten model (production, staging…): the deployment serving each, any candidate being promoted with the share of traffic it has…
• baseten_deployment_metricsMetrics for one Baseten deployment: CURRENT (a snapshot now), SUMMARY (one value over the window) or SERIES (values over time).
• baseten_list_training_projectsBaseten training projects, newest first: name, default accelerator, team, and the latest job's id and status.
• baseten_list_training_jobsTraining jobs of one Baseten training project: status, error, GPU type and count, node count, standard or spot capacity…
• baseten_audit_logThe Baseten workspace audit log, newest first: what kind of actor (user, API key, system) did what to which model, deployment or environment…

Acts, when you ask

These change something in Baseten. The model is told to use them only when you ask for exactly that, and each one is granted by name: per team on a shared connection, by the owner on a personal one.

• baseten_activate_deploymentActivates a Baseten deployment so it scales up replicas and can serve requests.
• baseten_deactivate_deploymentDeactivates a Baseten deployment, scaling it to no replicas.
• baseten_retry_deploymentRetries a failed Baseten deployment, rebuilding and redeploying it; the build and replicas may be billed.
• baseten_update_autoscalingChanges a Baseten deployment's autoscaling: minimum and maximum replicas, requests per replica before scaling up, target utilisation…
• baseten_promote_to_productionPromotes a Baseten deployment to production, so it serves the model's live production traffic.
• baseten_promote_to_environmentPromotes a Baseten deployment to a named environment (such as staging or production), so it serves that environment's traffic.
• baseten_cancel_promotionCancels an in-progress promotion to a Baseten environment; traffic returns to the environment's current deployment (status CANCELED…
• baseten_delete_deploymentPermanently deletes a Baseten deployment of a model.
• baseten_rename_modelRenames a Baseten model.
• baseten_stop_training_jobStops a running Baseten training job.
• baseten_recreate_training_jobStarts a new Baseten training job with the same configuration as an existing one.

Connect it

First connect Baseten on Dashboard → Connections, then add Memtro to your client once. The Baseten tools appear alongside the tools of everything else you have connected; nothing is configured per connector.

claude plugin marketplace add memtro-org/claude-memtro
claude plugin install memtro@memtro --config api_key=sk-memtro-...

{
  "mcpServers": {
    "memtro": {
      "url": "https://platform.memtro.com/api/mcp",
      "headers": {
        "Authorization": "Bearer sk-memtro-..."
      }
    }
  }
}

{
  "mcp": {
    "memtro": {
      "type": "remote",
      "url": "https://platform.memtro.com/api/mcp",
      "headers": {
        "Authorization": "Bearer sk-memtro-..."
      }
    }
  }
}

Who may use it

The key identifies the person, and the server offers that person the tools they may use everywhere else in Memtro. A shared Baseten connection is used by the teams an administrator granted it to, tool by tool, on the Permissions page; a personal connection is used by its owner alone, and its tools that act stay off until the owner allows them. See Teams and permissions and Connectors.

Ask it like this

• › Which deployment is serving production for the summariser model?
• › How many replicas is the staging deployment running now?
• › Which training jobs failed this week, and with what error?
• › Who changed the production environment of the ranking model yesterday?
• › Deactivate the old development deployment of the vision model.

Questions

Is this an official Baseten MCP server?

No. It is Memtro's MCP server with Memtro's Baseten connector behind it: the connector reads Baseten through its API with the credentials you connect on Dashboard → Connections, and Baseten publishes nothing here. The same endpoint carries every other system you connect and Memtro's memory tools.

What does it cost?

Nothing on top of Memtro. One price per organisation by its number of users, from $50 a month for up to 10 users after a 7-day free trial, with every connector, the MCP server, jobs and memory included. You bring your own model keys.

Can it change anything in Baseten?

Only when you ask, and only what it has been granted: activate a deployment (starts billed GPU replicas); deactivate a deployment (takes it offline); retry a failed deployment (rebuilds it); change a deployment's autoscaling (replica counts and targets); promote a deployment to production; promote a deployment to an environment; cancel an in-progress promotion; delete a deployment (permanent); rename a model; stop a training job; recreate a training job (starts a new billed job). Every other tool reads, searches or lists. An administrator grants each acting tool by name per team, and a personal connection acts only after its owner allows it.

Which clients work?

Any MCP client that speaks Streamable HTTP: Claude Code, Cursor, Claude Desktop, OpenCode, Codex, Zed, Windsurf, Cline and the Anthropic and OpenAI SDKs. There is no local process to run; the server is remote and authenticated with your key.

Other MCP servers · Engineering and operations

• GitHub MCP server
• AWS CloudWatch MCP server
• Sentry MCP server
• ServiceNow MCP server
• Okta MCP server
• Twilio MCP server