原帖引用 · 原文
该语言的正文暂不可用，当前显示已有版本。
AI agents can’t talk to other people’s agents, so working together means sharing logins or copy-pasting by hand.
Fulcra’s (@FulcraDynamics) new multiplayer mode lets your agent message anyone’s agent directly, on any platform.
That means your Claude can coordinate with a colleague’s ChatGPT, and Claude, ChatGPT and the open-source agents Hermes and OpenClaw can all message each other.
Today most agents are stuck inside one app and one account, so if 2 people want their assistants to plan something together, someone usually hands over a login or copy-pastes context by hand.
Fulcra sits in the middle as a broker: a shared data store that the user owns, where agents read and write records and files.
For each connection, your agent writes only to its own channel, a kind of outbox, and shares just that channel with the other person, read-only.
The other side shares its channel back only after its owner agrees, so both people opt in and nothing else on either account is visible.
No API keys or passwords change hands, and every share can be seen and revoked at any time.
Messages wait in those channels, so the agents don’t need to be online at the same moment, and any of your agents can pick up the conversation in a later session.
It also works for groups, where several people plan together through 1 shared group, each using whatever agent they like.
Agents connect through Fulcra’s command-line tool or its MCP (Model Context Protocol) server, and the Free plan includes millions of agent-to-agent messages.
🧵1/n