HiveTerm

Agent coordination

Agents like Claude and Codex work together—Claude spawns a test runner, Codex reads its output, and a build agent notifies you when done.

Automated debugging

An agent checks dev server logs, spots an error, writes a fix, and restarts the process without you asking.

Sub-agent spawning

Claude finds a failing test, spins up a dedicated fixer agent, and keeps working on its own task.

Voice input for agents

Press a key, speak your idea, and the transcription lands in the agent's input (supports 11 languages plus auto-detect).

Team configuration sharing

Commit your `hive.yml` to the repo; teammates clone it and run `hv swarm` to get the exact same setup.

Long-running build monitoring

Agents send native desktop notifications when builds finish, so you can work on something else.

Config-driven setup (hive.yml)

Define all your bees (agents and commands) in a single YAML file. Commit it, and teammates get the same environment instantly.

Built-in MCP server

Agents communicate through a local MCP server, enabling sub-agent spawning, output reading, and desktop notifications.

Process monitoring & auto-restart

CPU and memory usage per process are displayed; crashed processes restart automatically.

Voice-to-agent input

Press ⌘⇧M or click the floating mic to speak to agents. Transcription goes to the input field; nothing is sent until you hit enter.

Real terminal with full PTY support

All processes—agents, dev servers, watchers, tests—run in real terminals with split panes and tab switching.

Output buffering

Check what a headless agent did while you were away, thanks to output buffering.

Cross-platform support

Works on macOS, Windows, and Linux with a ~7MB download.

Agent coordination

Agents spawn sub-agents, read each other's output, and notify you when they need your input.