Skip to main content
hexgate chat is the inner-loop REPL — a single process against a local or registered agent. No platform, no Docker, no browser. It sets HEXGATE_LOCAL_MODE=1 automatically so audit stays on your machine even if HEXGATE_API_KEY lives in your .env from an earlier session. Denies and approval-required calls render as inline panels in the terminal — the same Decision data the platform would log, surfaced where you’re iterating. Reach for chat when you’re authoring a policy YAML, tweaking a tool, or shaping a system prompt. See Which path do I pick? for chat-vs-serve.
Run hexgate --help for all subcommands (chat, serve, register, policy) and hexgate <subcommand> --help for each one’s flags.

​
Agent resolution

--agent accepts either a plain agent id (example_agent) or a uvicorn-style module.path:attr spec (my_app.agents:my_agent). The CLI discovers agents from:
  • ./<agent_dir>/agent.yaml, ./agents/<agent_dir>/agent.yaml, ./examples/<agent_dir>/agent.yaml
  • code-defined agents registered from a --use script (e.g. examples/file_agents.py)

​
Keys the example agent uses

The example_agent (in the repo’s examples/example_agent/, resolved when you run from the repo root) uses the filesystem tools read_file / glob / grep, which need no API keys — only the model does:
  • OPENAI_API_KEY — the default openai:gpt-5.4 model
Bootstrap doesn’t hard-require it; it’s read when the agent is built (see the environment reference).