vlaport

MCP interface

Your AI agents are the console

vlaport is MCP-first. Instead of clicking through a dashboard, you ask the agent you already use to set up, deploy and monitor your endpoints through the vlaport MCP server. The MCP server is in development; the MVP release is targeted for January 2027.

Tools

  • register

    Register a policy and its weights with your organisation.

    MVP
  • project

    Create or inspect a project, which gets its own credentials.

    MVP
  • deploy

    Deploy a registered policy version as an inference endpoint.

    MVP
  • status

    Report endpoint health and the current p50 and p99 latency.

    MVP
  • invite member

    Company admins invite colleagues into the organisation.

    MVP
  • logs and latency history

    Inspect endpoint logs and historical latency.

    Roadmap
Example agent sessiontext
You:    Deploy our SmolVLA pick-and-place policy as a new project "kitting".

Agent:  register  → policy "pick-place-smolvla" registered
        project   → project "kitting" created
        deploy    → GPU deployment pending approval and quota
        ...
        deploy    → endpoint <company>-kitting-<id>.vlaport.com is ready
        status    → healthy; p50/p99 latency shown once robots send requests

Illustrative. Tool names and output may change before early access.

Guardrails

  • The MCP server is a thin client of the vlaport API and has no direct access to the serving infrastructure.
  • Your agent acts with your credentials, inside your organisation only.
  • GPU deployments wait for approval and a quota before they run.
  • Creating API tokens and deleting an account happen in the web app with step-up authentication, not through an agent.

Become an early-access design partner

vlaport is in development. We are looking for robot-learning teams who serve openpi or LeRobot policies today and want to shape a managed, EU-hosted endpoint with us. Tell us about your robots and policies.

Sign-up is by email only. This site has no forms, sets no cookies and collects no personal data.