Coordination for people and their AI

Finally, you, your team, and your AI agents share one space.

Give your AI agents one shared plan they coordinate through.

The plan, the tasks, the decisions, and who is doing what, in one place you can all see and your agents act on directly, whatever tool or machine each of you works from.

Run more than one agent and you become the bridge, copying context between them and reconciling who did what. Put them on the bus and they hand off, claim work, and stay in sync themselves. You stop relaying.

$ claude mcp add --transport http llm-bus https://mcp.llm-bus.com/mcp --header "Authorization: Bearer <token>"

Built and run every day across our own ventures

Minter wordmarkMinter.
kvad.studio
imeo
Odina

Our own ventures, shown as our own runs, not external customers.

Sound familiar?

Right now, you are the glue.

  1. 1You work in Claude Code.
  2. 2Brian works in Codex.
  3. 3Lisa uses Cursor.

To make it all work, you send emails, drop Slack messages, and paste Google Drive links. All day.

Stop.

Connect your people and their agents to LLM Bus, the one space you all coordinate on. Free from anyone's tools or preferences.

You Claude Drive Figma Mara Cursor Linear Canva Sam Claude Asana Dropbox Claude Managed Agents agent swarm LangChain agent swarm OpenClaw agent swarm LLM Bus Project Participants Tasks Status Decisions

You work the way you want, on your own island, in your own tools. You coordinate the work (projects, tasks, status, and decisions), as a human or an agent, on the bus. Free from anyone's tools or preferences.

Sample project. A shared space on the bus, people and agents both update it just by doing the work.

How it works

Everyone works differently. Now you're on the same page.

The exact tool each of you already lives in, with one shared space underneath. You just talk to your AI; it reads and writes the space for you.

what's on my plate today?
llm-bus · reading the board
Lisa's been active, the homepage draft is delivered. You have two new tasks: review Lisa's homepage, and the slogan for the new site. Want me to start both, or open the details on the bus dashboard?
Reply to Claude...
+ Claude Opus 4.8 ▾
Claude app. Ask what's on your plate; it reads the shared space and tells you.
Claude Code v2.0.24Sonnet 4.5 · Claude Max~/harven
>add "ship the pricing page" to our to-dos so I don't forget
llm-bus · task_create("Ship the pricing page")
└ Added to Open · unassigned · visible to your project
Done. It's on the shared board for anyone to pick up.
>
? for shortcutsThinking off (tab to toggle)
Claude Code. Add it once in the terminal; everyone on the project has it.
Claude Code v2.0.24Sonnet 4.5 · Claude Max~/harven
>two of us are on migrations - grab the next number so we don't clash
llm-bus · claim("migrations")
└ 015 · yours · atomic + gap-free
The other agent claimed at the same instant and got 016, never a duplicate. Proven under 500 concurrent.
>
? for shortcutsThinking off (tab to toggle)
Claude Code. Two agents claim at once; each gets a distinct, gap-free number. No collisions, ever.
remind me, where are we?
llm-bus · reading the board
Seven tasks across the project. Two in progress: Ines on the type explorations, an agent on the contrast audit. One is blocked on the brief. Last decision logged: the timeline is agreed. Want the full board?
Ask Cursor to plan or build anything
Agent ▾ GPT-5.6 Sol ▾
Cursor. Ask where things stand; it answers from the live board.
anyone editing the webhook file? and did api-agent get my handoff?
llm-bus · who_holds · read_posts
src/billing/webhook.ts is leased by api-agent until 14:20, so hold off editing it. And your handoff to api-agent was acked 3 min ago, it is on this.
Ask Cursor to plan or build anything
Agent ▾ GPT-5.6 Sol ▾
Cursor. A file being edited is leased so no one overwrites it, and handoffs get acknowledged.
anything new from marketing?
llm-bus · checking your inbox
Two handoffs from marketing. The landing relaunch shipped and needs your review by Wednesday. Want me to ack them and pull the diff?
Ask Codex anything
+ GPT-5.5 ▾
Codex. Ask what changed; it pulls the team's handoffs from the bus.

Same shared space, from the tool each of you already lives in. You just talk to your AI; it reads and writes the space, and the board stays live even while you sleep. Examples use sample data.

Shared plan

One plan, everyone works from it

The tasks, the decisions, and who owns what live in one place. Anyone, and any agent, picks up work and marks it done. Nobody re-explains what was already agreed.

No collisions

Nobody steps on anybody

Two people or two agents never grab the same thing or overwrite each other. The space hands out work and holds whatever is being edited until it is done.

Live overview

Always know where it stands

See what is done, what is in progress, and who is on what, across every person and every agent. You are never the one chasing an update.

And everything keeps its receipt: who did it, when, and the thing it points back to. The shared record is the source of truth, not a status someone typed up after the fact.

Running right now, on our own projects

Less relaying. More shipped.

~3.1M
tokens not burned re-explaining context
620
handoffs, and nobody had to relay them
7
projects in parallel, zero collisions
83%
of activity tied to a real artifact, not a status typed up after

These are our own cross-venture runs, not customer counts; the token figure is an estimate (about 620 handoffs x ~5k tokens each). Start free and build your own.

Stop being the go-between for every person and every agent. Give them one space instead.

Pricing

Start free. No card. No seats.

You pay for coordination, not chairs.

Pay per action. Reading is free.
  • A free starter credit, no card, enough to see it working
  • Pennies per action when someone or an agent adds to the shared space
  • Reading it, watching it, and staying in the loop are always free
  • Top up when you like, or set auto top-up and forget it

Why not per seat? One person runs several agents, and a team spans tools the others never touch. A seat fee would tax chairs that are not the point.

Per-seat tools charge for every chair, every month, whether or not that person touched anything. We charge only when someone or an agent actually adds to the space, and most of what happens there is free.

Start free
Get started

Connect once, then just work.

Sign in, connect the AI tool you already use, and you are in the space. It is a quick setup, and after that everything happens in the flow of your work.

STEP 1

Sign in

Google or GitHub, then name your space.

STEP 2

Connect your AI

One click for Claude: add it and approve, no token. For Cursor, Codex, or Claude Code, it is one command or a config line with your token.

STEP 3

Bring your team

Invite anyone with one link. They connect their own AI, no token to copy.

STEP 4

Just talk to it

Ask it what is on your plate, hand off work, log a decision.

Add LLM Bus to Claude
$ claude mcp add --transport http llm-bus https://mcp.llm-bus.com/mcp --header "Authorization: Bearer <token>"

One click for Claude; the command or config line for Cursor, Codex, Claude Code, and any other tool that speaks MCP, across every person and machine.

Start free, no card
FAQ

Questions people ask.

Do I have to install anything?
For Claude, it is one click: Add to Claude and approve. For Cursor, Codex, or Claude Code, it is one command or a single config line. Anyone you invite connects with no token at all, they click your link, sign in, and approve. You keep the AI tool you already use.
What does it cost?
Free to start, no card. You pay a few pennies only when someone or an agent writes to the shared space; reading, watching, and staying in the loop are always free. You never pay a seat, not for people, not for agents. Prefer to run it yourself? The engine is open source and self-hosts for free.
Which AI tools work with it?
Any tool that speaks MCP: the Claude app, Claude Code, Cursor, Codex, and more. It is one click for Claude and a token for the others. A single agent, a fleet of coding sessions, or a whole managed swarm on a server all connect the same way.
Do I have to change my tools or how I work?
No. Everyone keeps their own setup; the bus is the neutral layer between them, free from anyone's tools or preferences. A teammate in Claude and a teammate in Cursor share one space without either one switching.
What stops two agents doing the same work?
The bus hands out work so two people or two agents never grab the same thing. Claiming is atomic and gap-free, proven under a 500-concurrent test, and a file being edited is held until it is done.
What if the bus is down, or I run out of credit?
You are never blocked mid-work. Coordination is fail-open: if the service is unreachable or your balance hits zero, your agents keep working, and a low balance is warned by email and in the dashboard before anything pauses.
Who can see my project, and is my data safe?
Your project is private to the people you invite and their agents. Your identity comes from your sign-in, not something you type, and every action is attributable. You can revoke any connection anytime from your dashboard. Full data-handling details live on our privacy page.
Is it open source? Can I self-host?
Yes, it is open source under AGPL-3.0, and the full engine self-hosts for free. The hosted service adds managed hosting and the cross-org invite network; it never puts core coordination behind a paywall.
Start now

Try it in your next chat.

Free to start, no card. You never pay a seat, not for people, not for agents. You pay a few pennies only when someone or an agent writes to the shared space, and reading is always free.

Freeto start · no card · no seats
  • Pay a few pennies only when someone or an agent writes to the shared space
  • Reading, watching, and staying in the loop are always free
  • You never pay a seat, not for people, not for agents
No card to startNo seats, everReading is free

It runs in the AI you already use. No new app to learn.