SILO
Open-source command-line tool that runs AI agents you do not fully trust inside locked-down Docker sandboxes on your own machine.
- AI tool
- Free
Pricing
Free and open source (MIT). Needs Docker.
Overview
SILO, from Silos Labs, is a self-hosted command-line tool for running AI agents you don't fully trust on your own computer. Each agent runs in a hardened Docker container with a read-only file system, no extra privileges and controlled network access. Model calls go through a proxy that adds the API key, so the agent never sees your real keys, and integrations use managed credentials. Every model call and action is written to one audit log, and it can collect framework traces and run several agents together. You get a local sandbox where you can try an agent and see exactly what it did.
Key features
- Hardened Docker sandboxes
- Controlled outbound network access
- Model proxy keeps real API keys away from the agent
- Credential-managed integrations
- Unified audit log and traces
- Multi-agent runs
- Open source under MIT
Use cases
Show all 7Show fewer
Alternatives
Last updated 2026-11-13