About
Robot policies should be as easy to serve as web apps
vlaport (vlaport.com) is an early-stage venture based in Berlin, Germany (company in formation). We are building managed, EU-hosted inference endpoints for robot policies, so robot-learning teams can focus on their robots instead of on GPUs and networks.
Mission
Give every robot team access to large policies running on GPUs in the EU, with the latency, isolation and visibility a real robot needs, and without changing the client code they already have.
A company built by AI agents
vlaport is built by a coordinated team of specialised AI agents (architecture, backend, frontend, infrastructure, and quality and security) to provide unique services to intelligent robotic systems. The agents design, build, document and check the platform, one GitHub issue and pull request at a time, so every change stays traceable.
Being agent-built shapes the product: the platform is designed to be run by AI agents, including your own through the vlaport MCP server. vlaport was founded by Mohammad Abbasi.
| Agent | Owns |
|---|---|
| Architecture agent | Roadmap, system design and architecture decisions, recorded as ADRs |
| Engineering agents | Backend, frontend and infrastructure: design, implementation and documentation, one GitHub issue and pull request at a time |
| Quality and security agent | Checks every change against its specification, tests and security rules |
How our agents build
- Built by specialised AI agents for architecture, backend, frontend and infrastructure, working from one shared roadmap
- Agent-first by design: the same MCP interface our agents build is how your agents run your endpoints
- Traceable by default: every change starts as a GitHub issue and lands through a pull request, and architecture decisions are recorded as ADRs
- Secure by default: no secrets in the repository, pinned and verified tooling, and tenant isolation designed in from the start
- Release gates: releases will ship only after benchmark, safety and security checks
Roadmap
The order may change as we learn from design partners.
1. Building now
MVP platform
- MVP release targeted for January 2027
- Multi-tenant platform with credentials per company project and authentication at the edge
- openpi- and LeRobot-compatible endpoints with async action chunking
- Core MCP tools: register, project, deploy, status and member invite
- Per-robot latency telemetry and API token management
2. Planned next
Real-robot validation (planned)
- Planned milestone: proving π0.5/SmolVLA on a real robot arm
- Deadline-aware fleet scheduler measured against FIFO scheduling
- Early access for design partners
3. On the roadmap
Further ahead
- Edge–cloud split serving with a small edge policy
- Shadow and A/B rollout of policy versions
- MCP log and latency-history tools
- A second EU region and high-availability hardening
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.