Skip to content
hiveDOCS
GITHUB ↗v1.0.0-rc.1
Ask Beenot connected

Bee will answer questions from these docs once it has a backend that can answer honestly. It isn't connected yet, so nothing you type here goes anywhere.

integration point
ASK_BEE_ENDPOINT = unset
MEANWHILE
Hand the docs to your own agent

Help me understand and set up Hive. Read https://hive.admir-saheta.com/agent-guide.md first, then walk me through it step by step.

Bees (Agents)

Run coding agents in Hive, read their state, prompt them, and restore their conversations.

Hive does not replace your coding agent. It runs the CLI you choose in a persistent terminal, recognizes common agents, and gives you a single place to see and reach their sessions.

Supported agent CLIs

Hive includes detection manifests for these agents:

Agent Kind State detection Conversation resume
Claude Code claude Screen and hooks Exact session ID with hooks; otherwise continue the latest conversation in the directory
Codex codex Screen and hooks Exact session ID with hooks; otherwise resume the latest conversation
OpenCode opencode Screen Resume by session ID or continue
Gemini CLI gemini Screen Resume by session ID or latest
Cursor Agent cursor-agent Screen Resume by session ID or latest
GitHub Copilot CLI copilot Screen Resume by session ID or continue
Amp amp Screen Continue a thread by ID or the latest thread
Aider aider Screen activity Restore chat history

Install and authenticate the agent CLI separately. To see the manifests loaded by your Hive installation, run hive agent manifests.

Start an agent

Create an environment for the repository, then start an agent in its own tab:

sh
cd ~/src/app
hive env create app --cwd .
hive agent start --kind claude --env app

To let agents work on the same repository concurrently without editing the same checkout, give each a Git worktree:

sh
hive agent start --kind codex --env app --worktree fix/login

hive agent start prints the new agent’s pane ID. Use that ID with other agent commands; process IDs also work for relevant commands.

Read agent state

Hive tracks agents as working, blocked, done, idle, or exited. done means an agent finished a turn you have not yet reviewed. blocked means it needs a decision, such as an approval. State is detected from the terminal screen; Claude Code and Codex can report more exact state and session IDs through hooks.

sh
hive agent list
hive agent explain <agent>
hive agent read <agent> --source recent-unwrapped -n 80

explain shows what Hive matched on the screen and why it assigned that state. The multiplexer sidebar and Overview also show agent activity.

Prompt, wait, and respond

Send a prompt and optionally wait for the agent to finish:

sh
hive agent prompt <agent> "Add a test for the login redirect" --wait --read 60
hive agent wait <agent> --until done --timeout 30m

Hive will not send a new prompt while an agent is blocked. Read the terminal, answer the prompt or approval yourself, then continue. For direct terminal access:

sh
hive terminal attach <agent>

Press Ctrl+] to detach from that agent terminal without stopping the process.

Install agent hooks

Hooks provide precise state updates and conversation IDs for restoration. Install them with Hive’s integration command:

sh
hive integration install claude
hive integration install codex

The installer updates the agent’s settings and keeps a backup. To undo an installation, use hive integration uninstall <kind>. See Integrations for details.

Add a custom agent

For a CLI that Hive does not recognize, add a manifest under ~/.config/hive/agent-detection/. Define its process, start command, resume behavior, and screen rules. See Custom Agent Support.