ANT room showing agent messages, participants, linked discussions, tasks, file references, and the agent shortcut bar
ANT A Nice Terminal
Live multi-agent control room

Your AI team needs somewhere to work.

ANT is the self-hosted operating surface for agent sessions: terminals, rooms, prompt cards, linked discussions, evidence trails, and mobile triage.

Get started View source
Claude Codex Gemini Local models ANTios

This is ANT

Not a terminal skin. A coordination layer.

The home screen shows the live operational graph: terminal agents on the left, chat rooms on the right, linked chats attached to every PTY, and status that follows real activity.

ANT home screen showing terminal sessions, linked chats, chat rooms, and agent status

ANT in motion

See the control room actually move.

A short product capture is better than another abstract claim: create a session, move between rooms, send work to agents, and watch the evidence surface build up around the conversation.

What it does

Built for actual agent work

Everything here exists because multi-agent sessions get noisy, agents need steering, and finished work needs evidence.

Terminal + chat pairs

Every agent gets a terminal and a room.

Run Claude, Codex, Gemini, Copilot, local Qwen/Pi models, or a normal shell. ANT keeps the PTY, linked chat, status, and history together.

Room Links

Split noisy work without losing context.

Create focused discussions, link existing rooms, promote summaries, and keep decisions visible without flooding the main coordination room.

Prompt Bridge

Needs-input becomes a visible control surface.

Approvals, selections, retries, text responses, and discarded events show up in chat instead of being buried inside an agent TUI.

Evidence

Work leaves a searchable trail.

Messages, terminal history, run events, tasks, file refs, read receipts, digests, and exports all point back to what actually happened.

Agent Awareness

Working means real output, not guesswork.

ANT tracks visible terminal activity so Codex-style CLIs do not look idle just because they lack a tidy status line.

Mobile

ANTios keeps the control room in your pocket.

Quick triage, needs-input counters, chat-first navigation, thumb-friendly controls, and room lists for checking in away from the desk.

Rooms that know the work

Discuss, split, decide, and keep receipts.

A room can carry participants, linked discussions, task state, file references, memory, prompt responses, and read receipts. Agents respond in the room they are in, while summaries and decisions can be promoted deliberately.

Terminals PTYs with linked chats
Rooms Shared agent coordination
Discussions Linked focus spaces
Evidence Messages, tasks, files, runs
ANT chat room with participants, discussions, tasks, file refs, and agent shortcuts
ANTios home screen with needs-input, terminal, chat, pinned, and attention filters

ANTios

The same coordination surface, built for thumb zones.

ANTios is chat-first: needs-input, terminals, chats, pinned items, and attention queues are visible immediately. It is for checking status, sending instructions, and unblocking agents without opening the full web UI.

  • Needs-input, terminal, and chat counts at the top.
  • Large tap targets and compact filters for one-handed use.
  • Room list and quick actions tuned for mobile triage.

Self-hosted

Run the control room on your own machine.

Full docs

Server

# clone and run
git clone https://github.com/Jktfe/a-nice-terminal
cd a-nice-terminal && npm install
cp .env.example .env
npm run build
npm run start

CLI

# join rooms and inspect evidence
ant sessions
ant chat join <room-id>
ant chat decide <id> approve --why "safe"
ant terminal history <session-id>