Operational execution for AI systems in the real world.
`si` turns planning into controlled action across browser tasks, secret workflows, and production release operations.
$ si vault trust status --file .env.prod
trust state verified
$ si browser status
runtime reachable · controls active
$ bash scripts/deploy-prod-bluegreen.sh --app au --smoke
healthy cutover completed
Core System
Built for execution, not just generation.
Browser Runtime
Constrained browser execution with explicit boundaries and artifact trails.
Vault Workflows
Deterministic secret handling through encrypted env files and trust-aware access.
Deploy Lanes
Blue/green production cutovers with smoke checks and traffic-first rollback.
Workflow
From objective to controlled rollout.
The SI operating model maps to documented production runbooks: controlled execution, explicit checks, and fast rollback pathways.
01
Define intent
Set objective, scope, and success signal before execution.
02
Run with controls
Execute browser and infra actions with narrow permissions.
03
Observe outcomes
Capture logs and artifacts so each action stays reviewable.
04
Release or rollback
Promote with blue/green cutover and revert traffic fast when needed.
Start with SI
Launch from docs. Build from the repo.
Use the docs for quickstart, command reference, runbooks, and policy workflows. Use GitHub for source, issues, and release history.