Hardened sandbox
Dropped capabilities, read-only root filesystem, no new privileges and per-job network isolation.
Execution sandbox
RunFlow
Governed execution of containers over HTTP, with full lifecycle control.
Available
What it is
A REST API and MCP server for running Docker containers in hardened, isolated sandboxes on a dedicated host. Submit, pause, resume and observe jobs over HTTP; no SSH, no shared runners.
Who it is for. Teams that execute untrusted or machine-generated code and need isolation, observability and control rather than a general-purpose runner.
- Container execution
- DAG workflows
- MCP-native
- Sandboxed
- Pause and resume
What it does
Freeze and resume
Pause a live computation with its in-memory state and resume exactly where it stopped.
DAG workflows
Fan-out and fan-in pipelines with conditional branching, retries, data passing and human approval gates.
Live observability
Streaming logs, live CPU, memory and network use, and the full state history of every run.
Queue and schedule
Jobs queue with position and estimated start when at capacity; schedule for later or defer to business hours.
Every operation over MCP
Software agents connect with a URL and an API key; every operation is exposed as an MCP tool.
Ready to look at RunFlow?
The product site has the documentation and the way in. For procurement questions, a pilot or an integration, talk to us directly.