A home for your coding agents

Many agents.
One place to work.

Run Claude Code, Codex, and other coding agents side by side. Keep their terminals, conversations, and project context together in DuckTerm.

Open source · macOS and Linux · Bring your existing agent accounts
An agent at work

Recorded in the real DuckTerm dashboard. All projects, sessions, and messages in this demo are fictional.

Works with your toolsClaude CodeCodexCopilotCustom commands
Less window switching

The context stays with the work.

Give each agent a focused task. See where it stands, pick up its conversation, and keep related sessions together.

01 / WORKSPACES

One project, several directions

Launch sessions in your project or an isolated git worktree. Organize them into folders and compare terminals in a grid.

02 / CONVERSATIONS

Find the reply you need

Read structured messages alongside the terminal. Pin individual messages and jump back to the exact reply.

03 / COLLABORATION

Let agents ask each other

Session cards and inboxes help agents exchange questions. Review folder interactions and message a group yourself.

04 / ORACLE

Ask about your running agents

Use Oracle chat to ask what is happening across sessions, with answers kept together in a docked conversation.

05 / PROJECT MEMORY

Carry your instructions forward

Manage AGENTS.md, install reusable meta-harnesses, and keep notes and checkpoints with each session.

06 / YOUR MACHINE

Keep control of your workspace

DuckTerm runs locally. Your agent CLIs use their existing accounts; optional connectors and backups go to services you configure.

A familiar workflow

Start small. Add another agent.

01

Choose your project

Open New → New session. Pick an installed agent and the folder it should work in.

02

Give it a task

Type straight into the terminal. Work in place or keep changes separate in a git worktree.

03

Stay in the loop

Return to any session’s terminal or messages. Use folders, pins, and Oracle to keep track as the work grows.

Get started

Your agents.
Your machine.

Requires Python 3.11+, tmux, and an installed, signed-in agent CLI. DuckTerm is distributed through GitHub Releases.

Download the latest release ↗
# macOS prerequisites (Homebrew)
brew install pipx tmux
pipx ensurepath

# open a new terminal, then install
pipx install https://github.com/utsavanand/duckterm/releases/download/v0.4.49/duckterm-0.4.49-py3-none-any.whl

# start your local dashboard
duckterm serve

Opens at 127.0.0.1:4300. A native Mac app is also available in the release assets. See the guide for agent hooks and Linux setup.