Interactive Shell

ProbOS provides a Rich-powered interactive shell with natural-language input, 1:1 sessions with crew members, and 61 slash commands. Start it with probos (or uv run probos in a uv checkout); probos serve --interactive runs it alongside the API server and HXI.

The prompt shows the crew count and overall health:

[15 crew | health: 0.95] probos>

Natural Language

Just type what you want — the Ship's Computer decomposes your request into an intent DAG and executes it:

probos> read /tmp/test.txt                   # File read
probos> list the files in /home/user/docs    # Directory listing
probos> write hello to /tmp/out.txt          # Consensus-gated write
probos> search for *.py in /home/user        # Recursive file search
probos> what just happened?                  # Introspection
probos> why did you use file_reader?         # Self-explanation
probos> how healthy is the system?           # System health assessment

Talking to the Crew

Start a message with @ and a crew member's callsign to open a 1:1 session with that agent. The prompt changes to the agent's callsign, every following line goes to it, and another leading @callsign switches sessions. /bridge returns to the bridge.

probos> @Worf any security concerns this watch?
Worf ▸ what changed since yesterday?
Worf ▸ /bridge
probos>

Callsigns are chosen by the agents themselves at commissioning, so they differ from ship to ship; /manifest lists the crew aboard. An @callsign in the middle of a sentence is treated as a mention, not a session switch.

Slash Commands

/help lists every command with a one-line summary.

Status and Crew

Command	Description
/status	System status overview
/readiness	Ship readiness report
/agents	All agents with trust scores
/manifest	Crew manifest: /manifest [<dept>] [watch:<name>] or /manifest --ship
/ping	System uptime
/scaling	Pool scaling status
/credentials	Registered credentials and their status
/insights	Recent-activity summary: /insights [--days N] (default 7)
/debug	Toggle debug mode: /debug on\|off
/help	All available commands
/quit	Exit ProbOS

Memory and Knowledge

Command	Description
/memory	Working memory snapshot
/history	Recent episodic memory entries
/recall <query>	Semantic recall from episodic memory
/dream	Last dream report; /dream now triggers a cycle
/knowledge	Knowledge store status and history
/search	Search across all knowledge: /search [--type agents,skills] <query>
/rollback	Roll back a knowledge artifact: /rollback <type> <id>
/anomalies	Emergent behavior detection and system anomalies
/scout	Run a scout intelligence scan; /scout report shows the report

Plans and Feedback

Command	Description
/plan	Propose a plan before executing it: /plan <text>, /plan remove N, /plan
/approve	Execute the pending proposal
/reject	Discard the pending proposal
/feedback	Rate the last execution: /feedback good\|bad
/correct	Correct the last execution: /correct <what to fix>
/explain	Explain what happened in the last natural-language request

Orders and Authority

Command	Description
/orders	Standing Orders hierarchy and summaries
/order	Issue a directive: /order <agent_type> <text>
/directives	Active directives: /directives [agent_type]
/amend	Amend a directive: /amend <id> <new text>
/revoke	Revoke a directive: /revoke <id>
/imports	Manage allowed imports: /imports, /imports add <pkg>, /imports remove <pkg>
/conn	Delegate the conn: /conn <callsign>, /conn return, /conn status, /conn log
/night-orders	Guidance while the Captain is away: /night-orders <template> [ttl_hours], expire, status
/watch	Watch bill status
/alert	Bridge alert suppression: dismiss, resolve, mute, unmute, list
/grant	Clearance grants: issue, revoke, list
/tool-access	Tool permissions: grant, restrict, revoke, break-lock, list, check

Learning and Skills

Command	Description
/procedure	Procedure governance: list-pending, approve, reject, list-promoted
/gap	Capability gap reports: list, detail, check, summary
/qualify	Qualification tests: /qualify [run\|status\|agent <id>\|baselines]
/skill	Cognitive skills: list, discover, import, info, enrich, remove
/designed	Self-designed agent status
/qa	QA status for designed agents: /qa [agent_type]
/prune	Permanently remove an agent: /prune <agent_id>

Mesh Introspection

Command	Description
/weights	Hebbian connection weights
/gossip	Gossip protocol view
/attention	Attention queue and current focus
/cache	Workflow cache entries
/log	Recent event log entries: /log [category]
/federation	Federation status
/peers	Peer node models

LLM and Models

Command	Description
/models	Active LLM tier configuration: endpoints, models, status
/registry	All available models across tiers, the Copilot SDK and local providers
/tier	Switch LLM tier: /tier fast\|standard\|deep

Health and Diagnostics

Command	Description
/clinical	Clinical telemetry — dreams, traces, circuit breakers, audit (Captain authority)
/diagnostic	Multi-level system diagnostic: /diagnostic [<level>] [<focus>]

Scheduling and Voice

Command	Description
/remind	One-shot reminder from natural language: /remind <when> <what>
/schedule	Schedule from natural language: /schedule <NL>, list, cancel <id>
/wake-word	Custom wake-word trainer: status, collect, train, test

Sessions

Command	Description
/bridge	Return to the bridge from a 1:1 crew session