Open-source agentic runtime

Agents that do the work, on your terms.

Tonoman runs AI agents where your team already works. Bring your own compute, your own brain — a git repository — and your own skills. Run it yourself, free and open source, or let Tonoman Cloud run it.

Runtime
Open source
Inference
Your ChatGPT or Claude
Free on Cloud
1 agent · 3 people

Bring your own

Your compute, your brain, your skills.

Tonoman does not own your AI. It is the runtime that gives an agent a place to work, and the rules it works under.

Your compute

Run agents on your own machine or servers with the open-source runtime — as many as you like, free. Or put an always-on agent on Tonoman Cloud.

Your brain, in git

What an agent knows lives in a git repository you own: meeting recaps, notes, documents. Readable by people, versioned, and never locked in.

Your skills

A skill is what an agent can do — recap a meeting, watch for something, brief you on your day. Add as many as you like, and as many copies of each as you need.

How it works

One agent, set up once, answering in chat.

  1. Create an agent

    Give it a name and connect your own ChatGPT or Claude subscription. Tonoman never resells inference, so you never pay twice.

  2. Add skills

    Each skill brings the tools it needs. Add meeting-recap and it brings Plaud and your calendars; add it twice for a second set of logins.

  3. Give people access

    Only the people you give access reach the agent. Each connects their own logins privately, in a direct message.

  4. Work in chat

    Ask it in Slack, or let its skills run on their own: a recap filed after every meeting, a brief before the day starts.

Tonoman Cloud

Or let the cloud run it.

The same open-source runtime, hosted, with the Hub to manage every agent, skill and person from one screen.

Free to start

One always-on agent with unlimited skills, and three people. Agents on your own compute are free, however many.

A seat is a person or an agent

After the free seats, each person or always-on agent is a flat price a month. No per-message fees.

Add-ons when you want them

Transcription for your meetings, and a website your agents edit — each its own monthly price, cancel any time.

Questions

Do I pay for AI twice?
No. Every agent answers on your own ChatGPT or Claude subscription. Tonoman charges for seats and add-ons, never for the model.
Is it really open source?
Yes. The runtime that runs agents is open source. Run it on your own computer or servers without Tonoman Cloud at all.
Where does my data live?
What an agent knows lives in your own git repository. Your logins are kept sealed and are read only at the moment a tool uses them.
Which chat does it work in?
Slack today, with others on the way. Each agent is its own Slack app, generated and installed for you.

Start with one agent, free.

Unlimited skills, three people, and your own ChatGPT or Claude subscription.