---
name: help-inferenceos
description: Show all InferenceOS features, MCP tools, skills, and CLI commands available to the user.
user_invocable: true
---

# InferenceOS Help

Display a comprehensive overview of all InferenceOS features available in this session.

## Steps

1. Print the help card below directly — do not call any MCP tools.
2. If the InferenceOS MCP server is connected, mention that tools are live. If not, suggest running `claude mcp add inferenceos -- python3 -m inference_os --workspace . mcp-server`.

## Help Card

Print this in a code block:

```
╔══════════════════════════════════════════════════════════════════╗
║  InferenceOS — Local-First Agent Optimization Layer             ║
╠══════════════════════════════════════════════════════════════════╣
║                                                                  ║
║  SLASH COMMANDS                                                  ║
║  ─────────────                                                   ║
║  /help-inferenceos        This help screen                       ║
║  /inferenceos-dashboard   Visual token economics widget          ║
║  /inferenceos-local-brain Route tasks through local Gemma 4      ║
║  /inferenceos-token-saver Full token optimization workflow       ║
║  /inferenceos-context-briefing  Near-zero-token session start    ║
║  /inferenceos-task-manager      Persistent task tracking         ║
║                                                                  ║
║  KEY MCP TOOLS (call directly or let the agent use them)         ║
║  ──────────────────────────────────────────────────────          ║
║  Context & Search                                                ║
║  • inferenceos_route_task       Pre-search + route planning      ║
║  • inferenceos_search_context   Indexed workspace search         ║
║  • inferenceos_symbol_index     AST-powered symbol lookup        ║
║  • inferenceos_glob_files       Fast file pattern matching       ║
║  • inferenceos_grep_workspace   Regex search                     ║
║  • inferenceos_repo_snapshot    Live repo state                  ║
║                                                                  ║
║  Session & Memory                                                ║
║  • inferenceos_auto_store_turn  Persist conversation rounds      ║
║  • inferenceos_persist_result   Cache task results               ║
║  • inferenceos_refresh_session_note  Compact session summary     ║
║  • inferenceos_context_briefing      Workspace orientation       ║
║                                                                  ║
║  Token Economics                                                 ║
║  • inferenceos_dashboard        Visual savings dashboard         ║
║  • inferenceos_token_audit      Local vs naive cost comparison   ║
║  • inferenceos_budget_check     Pre-flight budget check          ║
║  • inferenceos_model_decision   Gemma/Claude router              ║
║                                                                  ║
║  Local Brain                                                     ║
║  • inferenceos_local_brain_think     Local pre-reasoning         ║
║  • inferenceos_local_brain_profiles  Switch model profiles       ║
║                                                                  ║
║  Domain Builders                                                 ║
║  • inferenceos_list_domain_builders  List specialists            ║
║  • inferenceos_activate_domain_builder  Get operating brief      ║
║  • inferenceos_create_domain_builder    Create new builder       ║
║  • inferenceos_capability_atlas         Full capability map      ║
║                                                                  ║
║  Research                                                        ║
║  • inferenceos_run_pico_experiment  PICO + Europe PMC search     ║
║  • inferenceos_match_workflows      Match task to workflows      ║
║  • inferenceos_fork_packets         Parallel task packets        ║
║                                                                  ║
║  Monitoring                                                      ║
║  • inferenceos_doctor           Readiness check                  ║
║  • inferenceos_circuit_status   Provider health                  ║
║  • inferenceos_analytics        Usage analytics                  ║
║                                                                  ║
║  CLI COMMANDS (run in terminal)                                  ║
║  ──────────────────────────────                                  ║
║  inferenceos serve              Local web UI                     ║
║  inferenceos mcp-server         Run MCP server (stdio)           ║
║  inferenceos doctor             Health check                     ║
║  inferenceos domain-builders    List domain specialists          ║
║  inferenceos brain-think "..."  Local pre-reasoning              ║
║  inferenceos analytics          Usage stats                      ║
║  inferenceos token-audit "..."  Cost comparison                  ║
║                                                                  ║
║  OPTIMIZATION PROFILES                                           ║
║  ─────────────────────                                           ║
║  standard    Balanced local compaction (default)                 ║
║  aggressive  Shortest context window, most compact               ║
║  zero-token  Local-only until escalation required                ║
║                                                                  ║
║  LOCAL BRAIN PROFILES                                            ║
║  ────────────────────                                            ║
║  gemma4-e4b-instruct   Advanced reasoning, 256K ctx (default)   ║
║  qwen3-4b-instruct     Fast routing on 16GB laptops             ║
║  llama32-3b-instruct   Minimal memory fallback                  ║
║                                                                  ║
║  DOCS: https://github.com/Reyanda/InferenceOS                   ║
╚══════════════════════════════════════════════════════════════════╝
```

Then below the card, add a one-line tip based on context:
- If starting a new session: "Tip: Run `/inferenceos-context-briefing` to orient to this workspace cheaply."
- If mid-session: "Tip: Run `/inferenceos-dashboard` to see your current token savings."
- If the MCP server is not connected: "Tip: Run `claude mcp add inferenceos -- python3 -m inference_os --workspace . mcp-server` to connect."