Transparent agentic automation you can actually audit. Every skill is a single Python file. Every execution is logged. Every action is observable.
Single Python files in .boots/skills/. Each declares a manifest block describing its name, triggers, permissions, and resource needs. The engine audits every skill before loading it.
Every skill is scanned: static analysis for dangerous imports, dependency vulnerability checks via OSV, and behavioral analysis. Skills earn trust tiers: unreviewed → sandboxed → verified.
Your AI-powered operator interface. The Concierge manages personas, sessions, and memory. Talk to it in the Concierge panel — it can invoke skills, recall past context, and observe your workspace.
Configuration profiles scaled to your hardware. micro_g for CPU-only machines, earth_g for cloud API access, jupiter_g for GPU clusters. Detected automatically at boots init.
Every skill execution produces a structured "nutrition label" — token usage, cost, latency, and a full execution trace. Right-click any skill → Show Nutrition Label.
| Location | Purpose |
|---|---|
.boots/ |
Workspace root — config, databases, skills, and logs |
.boots/skills/ |
Installed boot skills (one .py file each) |
.boots/config.json |
Workspace configuration — tier, LLM, daemon settings |
.boots/Personas/ |
Concierge persona definitions (markdown files) |
.boots/memory.db |
Persistent memory store — semantic, episodic, procedural |
.boots/state.db |
Execution state — skill runs, approvals, results |
| Command | Action |
|---|---|
Boots: System Panel |
Open the System Panel — live status, services, actions |
Boots: Run Skill |
Execute a skill by name |
Boots: Genome Scan |
Describe what you want — AI generates a skill for it |
Boots: Switch Persona |
Change the active Concierge persona |
boots init |
Initialize a workspace (CLI) |
boots daemon start |
Start the daemon manually (CLI) |