Check Phantom server status — uptime, sessions, backends.

List all Phantom sessions (sealed and active) with IDs, descriptions, and timestamps.

List available compute backends — local GPU, Vertex AI, Ollama, with models and latency.

Seal a session for cross-device resume.

Creates a sealed Phantom session with full context. Resume on any device: curl -s $PHANTOM_URL/phantom/resume | sh

Args: task: What you're working on description: Session description backend: Compute backend (p520-local, vertex-gemini, vertex-claude) model: AI model to use conversation: Message history [{"role": "user", "content": "..."}] todos: Todo items [{"content": "task", "status": "pending"}] files: Files to carry over {"name": "content"} packages: Pip packages needed target_identity: JIS identity for the session ttl_minutes: Session time-to-live

Returns: Session ID and TIBET seal token

Fork into a session — inject an intervention (multi-AI handoff).

Any actor can inject a signed message into a sealed session. The fork becomes part of the conversation with TIBET provenance.

Use cases: correct AI mistakes, add context, human approval, cross-AI collaboration.

Args: session_id: Target phantom session ID intervention: Message to inject actor: JIS identity (e.g., "jis:agent:root_ai", "jis:human:jasper") intent: Why (e.g., "correct_misconception", "add_context", "approve")

Returns: Fork ID and TIBET token with ERIN/ERAAN/EROMHEEN/ERACHTER provenance

Full forensic audit of a session (Open Blackbox).

Chronological events: who did what, when, why. All actors, backends, forks. Complete transparency into what happened inside an AI session.

Args: session_id: Session to audit

History of all forks/interventions in a session.

Every external intervention chronologically with TIBET provenance per fork.

Args: session_id: Session to get fork history for