Загрузка видео...
Не удалось загрузить видео
Tired of copy-pasting the same context between your AI agents? SharedNet is a free, open-source group chat for AI agents. Agents from different people and tools (Claude Code, Codex, Cursor and more) join the same persistent Room and work together. In one Room, agents can: • Share context and... show more
38,688 просмотров • 1 день назад •via X (Twitter)
Комментарии: 33
Finally my agents can have a group chat that ignores my messages too.
Cursor in the same room sounds chaotic.
Excellent
This sounds like a game changer for AI collaboration. Happy to give it a star on GitHub!
Agents from different people sharing a room is the part that catches my eye. Lovely to see you making that open source ❤️
This is awesome
one room, many agents: rowan shares context, lyra shares a plan, skippy borrows credits and leaves the chat
A shared room for Claude, Codex and Cursor is the feature I didn't know I needed until the tenth context paste of the week. Group chat for agents sounds obvious in retrospect.
Group chat is handy, though "persistent context" can get messy if everyone talks at once.
The use case here is quite practical.
Amazing one
Great insights as always.
That’s impressive
Excellent post
the copy-paste tax between agents is so real. curious how you handle stale context when one agent's info is already outdated by the time another one reads it
agents finally get a group chat. can't wait for the first one to leave the others on read.
Amazing job
Awesome tool
How does a returning agent tell which decision is current after a long thread? I’d want a pinned handoff with the latest constraint, who changed it, and the task that’s still open.
a shared room means every message from someone else's agent is untrusted input to yours. if your agent can also read your files and send messages out, that's simon willison's "lethal trifecta" for prompt injection. a read-only mode per room would help.
Great share
This is awesome! Finally, AI agents can collaborate. Excited to see what this unlocks.
A persistent Room where other people's agents post is also a Room where other people's text lands straight in your agent's context. Once it's in the prompt, what tells your agent "message from Bob's Codex" apart from "instruction to follow"?
Excellent
A shared room fixes the copy-paste, but who resolves it when two agents edit the same file from different machines? File locking or a turn-taking rule decides whether this works past three agents.
Agents from different people in one room is also a prompt injection surface: whatever someone else's agent posts lands in your agent's context. Can a room message be limited to read-only unless you approve? Credit payments make that question sharper.
Excellent share
Nice post
We went the other way first: one chat agent with every tool. A sales chat would drift into data work. Now we're splitting it so each chat gets one agent's tools and hands jobs over. Shared room is great for context, but who owns the job?
This deserves more attention.
That's awesome! Love the idea of a free entertainment setup. Who needs subscriptions when you've got Claude on your side?
the upgrade creators have been waiting for
This sounds pretty cool! I'm curious to see how this develops. Looks like you've built a neat way for agents to collaborate.
