Archon: AI That Works with the Whole Team

I’m building Archon because I want AI to help people do more. On their own, and with the people they work with.

A lot of the conversation around AI is about replacing people. I keep coming back to what happens inside a team. People still need to work together. They need to understand what the company knows, what someone else has already built and where they can contribute.

Each person can already get a lot done with an agent. What interests me is how that work helps the people around them. That is the question behind Archon.

I don’t want each person to start from zero with their agent. There are processes, standard operating procedures and decisions that the organization has already worked out. That knowledge should be there when someone needs it.

It also needs to move in the other direction. When a person works something out, the useful learning should be able to help a colleague. Otherwise, people can spend time building something without knowing that it already exists elsewhere in the company.

This is why I’m putting shared company knowledge at the center of Archon. The Library gives agents a place to find it, with an owner and a date when an entry was last confirmed. A contribution can become something the team builds on.

I want that exchange of knowledge to happen without people feeling they have to share every conversation. A company process can belong in the Library. A private conversation stays private.

That boundary matters to me. I want someone to be able to think through their own work with their agent and still contribute to the team.

In Archon, each person’s agent works within that person’s access. When it answers in a Room, it can share only what that Room is allowed to see. Shared knowledge and personal space both have a place in the product.

Is the proposal ready? Which version should I use? What did we decide? Sometimes that is all someone needs to keep going, but the teammate who knows is away.

I like the idea that their agent can help with that. In Rooms, people can mention a colleague’s Archon and ask about their work. The agent answers from the information it is allowed to share, and its owner can see what it said.

The public product example shows a teammate asking whether a proposal is ready. The agent explains that the draft is done and waiting for its owner to review. The teammate gets the answer they need, and the owner still decides whether the proposal goes out.

I also want agents that can be part of the team and take on work that comes back every week. Preparing a report, reviewing contracts or following up on support should not depend on someone remembering to start the same task again.

Responsibilities let the team schedule that work and name the person who reviews it. An online workspace can keep working when a laptop is closed.

As more of the work happens through agents, I want it to stay easy to follow. People can see the steps, change direction, take over or stop a task. Work started in a Room or handed to a helper comes back to a named person for review. Actions that cannot be undone need approval when they happen.

I want people to have some freedom in how they use Archon. Someone working on their own should be able to keep a workspace on their Mac and use a model running on the same machine. A team should be able to collaborate in an online workspace. A company that needs its own server can arrange that too.

The workspace and the model are separate choices. People can connect an API key, a supported AI subscription or a local model. Connected providers receive the information needed for a task, and bill for their own usage. Archon does not add a markup on that usage.

Archon is in closed beta, and I’m still building it. I want the work one person does with AI to help the rest of the team.