Tachyonic Documentation
Sandboxes and security runtimes for AI agents
Tachyonic runs agent-written code in sandboxes and tests authorized agent endpoints with security runtimes. Use one customer account, workspace keys, CLI, and API origin. A sandbox runs a job; a runtime tests how an agent behaves.
$ curl -fsSL https://tachyonic.sh/install | bash$ tachyonic login$ tachyonic sandbox create --budget 30m$ tachyonic sandbox exec <sandbox-id> -- python3 --version$ tachyonic sandbox rm <sandbox-id>$ Sandbox commands require CLI 0.18.0 or later. The installer serves the latest
published release; check tachyonic --version before following the sandbox guide.
Container isolation shares the host kernel. Choose the image, limits, and
permitted hosts deliberately.
Start here
Sandbox quickstart
Install, log in, and run your first sandbox command.
Sandboxes
Isolation, network policy, budgets, files, and usage.
Runtime quickstart
Test an authorized agent endpoint and inspect its evidence.
Runtimes
Lifecycle, limits, regions, budgets and approval gates.
Verify evidence
Check the signature on an evidence bundle locally.
CLI reference
The public commands and their authentication contract.
Reference
Sandbox CLI
Sandbox and workspace commands, flags, and exit codes.
Sandbox API
Workspace-authenticated sandbox and account routes.
Attack catalog
Categories, OWASP 2026 and ATLAS relations, severity and scoring.
Platform API
REST endpoints for targets, scans and findings.
CI/CD
Run scans in pipelines and fail builds on findings.
Configuration
Scanner configuration options.
The default runtime path supports signed evidence. Regional signing, retention, and sharing must be checked for the path you use. Download the exact bundle and manifest, and verify them with a trusted public key before relying on them.