JarvisCore 2.3: multiplayer agents. Your team and its agents in one thread

Hey makers 👋

Today we've shipped multiplayer in JarvisCore.

Most agents are built for one person in one chat. But work happens in teams: people disagree, change their minds, and add "oh, and also…" while the work is still running. Multiplayer is built for that.

How it works:

• A room holds the team's shared files and who can do what (owner, editor, viewer).

• Anyone asks in their own words. The agents' work comes back as a proposal you can read as a diff or preview, then accept or turn down.

• When two people push back at the same moment, one revision runs and both voices are kept. Nothing anyone adds while work is running is lost.

• Only the person making a decision gets the Confirm button. Agents propose; people decide.

• Agents act with the access of whoever asked. Work that used someone's private data is shown to them first.

• It runs where your team already talks: Slack out of the box, or your own product's chat.

Watch example here: https://youtu.be/AAFECpBe-VQ

Try it:

pip install -U jarviscore-framework

Guide: https://jarviscore.developers.prescottdata.io/guides/multiplayer-agents/

GitHub: https://github.com/Prescott-Data/jarviscore-framework

We'd love your input:

1. Where would your team want agents in the room: Slack, your product, somewhere else?

2. What's the first team workflow you'd hand to multiplayer agents?

3. What would you need to trust agents working on shared files?

I'll be in the comments all day. 🙏