v0.0.33 · a tribute fork · installs beside the real thing

The command center
was missing a commander.
Meet T3 Agent Orchestrator.

T3 Code already runs every coding agent you own: Claude, Codex, Cursor, side by side in one inbox. The glaring hole: no native agent of its own, trained to orchestrate that fleet on your behalf. That's what this is. Native, built in, running out of the box on a single API key, on your model, or entirely on your own machine. And it runs the rest of them for you.

macOS, Apple Silicon and Intel · One API key, running out of the box · Basically a mini Theo · Unsigned preview, two clicks past Gatekeeper · MIT, like upstream

Watch it run

It narrates its own demo.

One prompt. The agent fans six questions out across Claude and Codex, reads the answers back, follows up, files the threads, and announces every move out loud through the Mac's speech engine while it works. Nobody touches the keyboard after Enter.

100 seconds, one take· six live threads across two providers· the voice is say, run by the agent itself· the whole thing ran on DeepSeek V4 Flash for under a penny


The headline act

One agent that runs
all your agents.

T3 Agent isn't another entry in the provider list. It operates the list. It knows which agents you have, what each costs, and what each is good at. You say what you want done; it decides who does it, launches them in parallel, and reads their work back.

you ›Have Codex fix the flaky auth test while Claude reviews my open PR. Separate worktrees, and wake me when both settle.
you ›Fan this rename out to three cheap threads in parallel, then read back what each one actually changed.
you ›My Claude thread is stuck on a question. Answer it and let it keep going.
T3 Code
Search
Better T3 now
Auth PR review
Better T3 now
Flaky test fix
Settled
delegate_to_agent Claude Code · review the auth PR · own worktree delegate_to_agent Codex · fix the flaky test · own worktree wait_for_thread ×2 · a real wait, not a poll
Both settled. Claude flagged two issues on the PR; Codex's fix is green on t3-agent/flaky-test. Nothing merged without you.
Ask for follow-up changes or attach images
T DeepSeek V4 Flash Default Full access

the app's own dark theme, tokens and all·on loop·the real thing is the video above

It knows what you're paying for

It can see which providers are configured, whether each one bills per token or draws on a subscription you've already bought, and exactly which models each will accept. So it can put the cheap work on the plan you're already paying for.

It picks the dial settings

Provider, model, and how hard to think, chosen per task. Ask for a reasoning level a model doesn't offer and it uses the nearest one that model does, instead of refusing and wasting the round trip.

It reads what actually happened

Not the summary a thread wrote about itself. The real transcript, with the failed tool calls marked and the errors counted. If a delegate is stuck waiting on a question, it can see the question and answer it.

It cleans up after itself

Send a correction back into the same thread, rename it, snooze it, pin it, settle it or archive it. The inbox stays yours, but you don't have to do the filing.

Delegations can't collide

Every delegated thread gets its own git worktree and branch, so five agents can work the same repo at once without overwriting each other. You merge what you like, when you like.

It waits without burning tokens

Tell it to wait on a delegate and it holds on a real server-side wait (no polling loop) and wakes the moment the thread settles or stops to ask a question.

Before

  • Open the sidebar, find a thread
  • Type the task, pick the model, wait
  • Read it, decide if it worked
  • Open the next thread
  • Type roughly the same thing again
  • Repeat, per agent, per task

You are the message bus.

After

  • Say what you want done
  • It picks who runs each piece, and on what
  • It starts them all
  • You ask how it's going
  • It reads the real transcripts back
  • It fixes, answers and archives

You supervise a swarm from one chat.

Also in the box

A real coding agent
in its own right.

A dispatcher is only as good as its own hands. Before T3 Agent orchestrates anything, it does the ordinary job properly. Not a chat box bolted onto the sidebar, but a first-class in-process provider, so every feature T3 Code already has works on it unchanged. Paste one API key and it's coding straight away, on models priced like DeepSeek V4 Flash, the one the demo runs on.

It writes code

Reads, writes, edits, globs, greps and runs commands, with path safety that keeps every one of those inside the project.

It asks permission

Obeys the thread's runtime mode and the same approval prompts as every other provider. The gate is a pure function, tested on its own.

It reads your rules

Picks up AGENTS.md and CLAUDE.md, loads skills, and connects the MCP servers you've configured.

It checkpoints

Every turn is a hidden git ref, so you can rewind a thread exactly the way you already do with Codex or Claude.

It spawns sub-agents

Fans work out to parallel sub-agents when a task branches, and compacts the conversation when it gets long.

It shows you the bill while it runs

Every thread carries a live cost readout: total spend and what the last turn cost, priced from your provider's own published rates, cache discounts included. A model it can't price reads "unknown" instead of pretending to be free.

Try it, right here

It sits in the picker like
it was always there.

The picker, exactly as it ships, in dark. The whole frontier filed under one provider: yours.

Ask for follow-up changes or attach images
T DeepSeek V4 Flash Lat… Default Full access
T3 AgentT
Search models...
Kimi K3
T T3 Agent · Moonshot AI
⌘1
DeepSeek V4 Pro
T T3 Agent
⌘2
Qwen3.8 Max
T T3 Agent · Qwen
⌘3
MiniMax M2.7
T T3 Agent
⌘4
GLM 5.2
T T3 Agent · Z.ai
⌘5

Bring one key

One OpenRouter key.
The whole frontier.

OpenRouterThe default, and the path this preview was tested hard on. One key, four hundred models behind it, prompt caching marked on every request, exact per-model pricing read live from their catalogue. Paste the key and go.
recommended
It speaks three other backends too, if you need them
AnthropicStraight to the source, on your own API key.
direct
CerebrasFor when you want the tokens to arrive absurdly fast.
direct
Anything OpenAI-compatibleOllama, LM Studio, vLLM, your own box. Fully local, no key required, nothing leaves the machine.
no key needed

These ship and work; they just have fewer miles on them than OpenRouter does.


Under the hood

Everything it can do,
in one list.

No hidden capabilities and no marketing verbs. This is the actual tool surface.

Working on your code

read write edit glob grep bash webfetch update_plan sub-agents skills mcp servers

Running your other agents

list_providers list_models list_projects list_threads create_project delegate_to_agent wait_for_thread read_delegated_thread send_to_thread answer_thread_question approve_delegated_request stop_delegated_thread revert_delegated_thread set_thread_state rename_thread

Read this part

What it won't do.

Stated up front, because finding out later is worse.

01
It waits when told to. It doesn't lurk.

It can hold on a delegated thread until the work settles: a real server-side wait, not a polling loop. But nothing pings it unprompted: there's no background watcher, and it will tell you that rather than promise to watch things it can't see.

02
It can't delete anything.

Not threads, not projects. Deletion is the one action with no undo, so it isn't wired up.

03
It can't change what another thread is allowed to do.

Which means it can never widen its own reach by raising a delegate's permissions.

04
It won't answer approval prompts for you.

Not unless you turn that on separately. Approvals are how you stay in the loop.

05
Delegating spends the target provider's credits.

Nothing per turn on a subscription; real money on an API key. There's no cap on how many run at once, so a wide fan-out spends widely.

06
It's early, and it's a fork.

Upstream says to expect bugs. This adds twenty-two thousand lines of code on top of that. Treat it accordingly.

Get it

Running in about
ninety seconds.

macOS · Apple Silicon & Intel

Installs as T3 Code (T3 Agent), with its own name and its own bundle ID, so it lands beside your real T3 Code, never on top of it.

Download .dmg
arm64 and x64 · ~150 MB each v0.0.33 unsigned preview, macOS objects once, steps below
  1. Grab the DMG for your Mac: arm64 for Apple Silicon, x64 for Intel ( → About This Mac if unsure). Open it, drag to Applications. Your existing T3 Code is untouched.
  2. First launch only: macOS balks at unsigned apps. System Settings → Privacy & Security, scroll down, Open Anyway. Signed builds are coming; this is the preview.
  3. In the app: Settings → Providers → +, pick T3 Orchestrator, T3 Agent's name in the provider picker.
  4. It defaults to OpenRouter. Paste your key into the field marked sensitive, which keeps it out of settings.json.
  5. Start a thread, pick a model, go.

One caveat worth knowing: it shares its on-disk data (~/.t3) with an official install: same projects, same threads, same provider keys. Run one at a time.

Windows or Linux

No prebuilt binary yet; build it from source. Needs Node 22.16+, 23.11+ or 24.10+.

# clone the branch
git clone -b feat/t3agent-upstream-pr \
  https://github.com/BenSyne/t3code.git
cd t3code

# install vp, skip if you have it
curl -fsSL https://vite.plus | bash

vp i
vp dev

One thing to watch for

npx t3@latest installs upstream, which does not include the agent. The only ways to get this build are the download above or the source clone beside it.

Everything upstream does, this still does. It's upstream with one provider added, not a rewrite.


Dear T3 team

No PR. You're welcome.

First, the important part: this is a tribute. T3 Code is my favorite tool in the category, the best multi-agent command center anyone has shipped, and I use it every single day. I built this because I love the product, not because I think you're doing it wrong.

You've said thousand-line PRs are a good way to get banned. This one is twenty-two thousand, so it lives on my fork: full diff, comments open, nobody's inbox harmed. Your README says to fork the editor you want. So I did. Anything useful is MIT: take it.

171files changed
21.8klines of code added
85commits, each standalone
395tests on the agent alone

Built on upstream 89c320df · v0.0.33 · 11 Aug 2026