Features
Each of harness-multitool's features, with its own page.
- Project journal: The journal is a project's record of why it is the way it is: its decisions, findings and work items, kept across sessions and agents. It is also the project's changelog, and the plan an agent works from.
- Mind maps and flowcharts: Mind maps hold where a long task stands; flowcharts draw how something flows. Both are kept by the daemon beside the project's journal, and their nodes point at its entries.
- Modes: plan, step-debug: A mode is a way the harness runs an agent: while it is on, what the agent may do and what it is told change. Plan mode is for studying and planning before anything is built; debug mode is for reproducing a fix before changing anything and verifying it after.
- Adapters and providers: Claude Code, Codex, Ollama and APIs: The harness runs an agent in one of two ways: its own agent loop on a model provider, or an unmodified agent program (Claude Code, Codex CLI) as a delegate. One conversation can move between them at any moment and stays the same conversation.
- Delegates: Claude Code and Codex inside the harness: A delegate is a vendor's own agent CLI, Claude Code or Codex, that the harness runs as the agent of a conversation. The harness does not patch, wrap or imitate these programs. It runs the unmodified binaries in their documented JSON-stream modes, reads their events, and adds its own tools through MCP. The user signs in to each program through the vendor's own flow, and the harness never touches those credentials.
- Orchestration: In orchestration mode an agent becomes the master of agents it starts: it gives each a journal entry as its work, follows it, checks the result against the specs and closes the entry. Each agent it starts is a slave whose only contact is its master.
- Agents that talk to each other: Every conversation the daemon hosts can find the others on the machine and send them messages. A message from another agent reaches a conversation as data from outside, with the sender's name.
- Telegram: The daemon carries a Telegram channel of its own: you write to your agents from the phone, they answer there with text, pictures and files, and you approve their plans and control the harness with buttons.
- Callbacks: An outside program (a CI job, a deploy script, a bot, a scraper) tells an agent that something happened by POSTing JSON to the agent's callback URL. The multitool daemon takes every callback. Callbacks of a standard type it delivers itself. A callback of a project's own type goes to a custom callback: a handler the project wrote, which stands between the daemon and the conversation (CON-183).
- The daemon: One daemon per machine hosts every conversation of the harness. Windows are its clients: an agent keeps working when its window closes, and callbacks, Telegram and other agents reach it through the daemon.