Development computers for software agents.
Give every coding agent a complete development environment on demand — with the compute, repositories, storage, credentials, and tools it needs to build and test real software.
Built by Semaphore, running software-development workloads at scale for more than a decade.
The missing execution layer
Agents need more than a sandbox.
Coding agents can reason about code and decide what to do next. To build real software, they still need somewhere to work.
They need to clone repositories, install dependencies, compile code, run tests, start services, use credentials, and persist important data.
Your software factory handles the intelligence. Semaphore handles the development infrastructure.
POST /computer → software development can begin
Computer lifecycle
From API call to development environment.
REST-first from creation to deletion. Hosted MCP will expose the same infrastructure directly to agents.
POST /computersChoose a machine
Start an F1 x86 development machine for the task.
Attach context
Mount Storage, check out repositories, and provide task secrets.
Develop software
Run commands. Build. Test. Debug. Start services.
Persist and delete
Keep what matters in Storage. Delete the Computer when work is done.
Bring the software factory. We’ll provide where its work runs.
Request early accessDevelopment-ready by design
A development computer, not an empty VM.
The goal is not merely to boot Linux. It is to get an agent productive on real software.
Full development machines
Semaphore-maintained x86 environments instead of a minimal runtime container.
Multiple repositories
Check out every repository a task needs before the agent begins work.
Persistent Storage
Keep durable data in Unified Semaphore Storage while Computers stay disposable.
Secrets & identity
Use Organizations, Service Accounts, RBAC, and Secrets to control access.
REST-first
Automate the complete Computer lifecycle through an API built for software factories.
Hosted MCP
Let MCP-compatible agents operate Computers without installing a Semaphore binary.
Durable context
Computers come and go.
Your data doesn’t have to.
Compute is ephemeral. Storage is durable. Continue later on a fresh Computer without keeping an idle machine alive.
Repositories · dependencies · artifacts · development state
Built for concurrency
Infrastructure for one agent—or a factory of them.
A software factory may need one development environment today and many running in parallel tomorrow. Semaphore.Computer is designed for programmatic creation and high-concurrency software-development workloads.
Security foundation
Autonomous agents.
Explicit boundaries.
Software agents can execute code, access networks, and use credentials. Semaphore.Computer builds on Semaphore’s existing identity and administration foundation.
- Organization-scoped access
- RBAC and Service Accounts
- Managed Secrets
- Isolated development machines
- Durable Organization Storage
- Metered resource lifecycle
Predictable spend
Give agents a budget, not a blank check.
Buy prepaid credits first. Let your software factory consume them while compute and Storage usage are metered against an explicit budget.
Agents can consume budget.
They can’t create budget.
Semaphore pedigree
New product.
Proven infrastructure.
For more than a decade, Semaphore has provisioned isolated development machines, run builds and tests at high concurrency, managed source-code access and secrets, and metered compute and storage.
Semaphore.Computer turns that foundation into a standalone development-compute platform for software agents.
Initial product scope
What we’re building first.
Focused infrastructure primitives for agents and the systems orchestrating them.
We are not building another coding agent or agent orchestrator. Bring the agents and software factory you already use.
Early access
Building a
software factory?
We’re working with teams building coding agents, internal software factories, and AI-native development platforms. Request early access to help shape the API and development environment.