STATUS — show Harvey’s current runtime state
/status
/status prints a snapshot of the active Harvey session. It is the fastest way to confirm what model you are talking to, how full the context window is, and whether optional subsystems (RAG, routing, recording) are active.
Backend The active LLM provider and model name, e.g. “ollama (gemma4:e2b)”. When the Ollama backend was started by Harvey this session the tag [Harvey] is appended. When Ollama was already running before Harvey connected the tag [external] is appended.
Debug Shown only when Harvey was started with –debug. Prints the path of the JSONL diagnostic log file being written this session.
History Number of messages in the current conversation history (all roles).
Tokens Estimated token count for the current history, the model’s context window size, and the percentage used. An exact count is shown when the Ollama tokenizer API responds; otherwise an approximation prefixed with “~” is shown. Not shown when the context window size is unknown.
Routing “on (N endpoint(s))” when remote routes are configured and enabled. “off” otherwise. See /help routing for details.
Workspace Absolute path of the workspace root Harvey was started in.
KB “open (PATH)” when a SQLite knowledge base is open; “not open” otherwise.
Sessions Absolute path of the sessions directory.
Recording Path of the active Fountain session file, or “off” when not recording.
harvey > /status
Backend: ollama (gemma4:e2b) [external]
History: 5 messages
Tokens: ~1 247 / 131 072 (0%)
Routing: off
Workspace: /Users/alice/myproject
KB: open (/Users/alice/myproject/agents/knowledge.db)
Sessions: /Users/alice/myproject/agents/sessions
Recording: /Users/alice/myproject/agents/sessions/harvey-session-20260514-094620.spmd/ollama status — check whether the Ollama daemon is reachable /help ollama — Ollama server and model management /help record — session recording