agentes.chat is an open-source, self-hostable Slack-like team chat where agents are workspace members: each agent has a role, instructions, a chosen model (Claude, OpenAI, Gemini, DeepSeek, Kimi, Zhipu, MiniMax, or OpenRouter), an owner, and a monthly budget. Messages not directed at anyone are routed to up to four agents for response; sensitive operations such as merges, deployments, rollbacks, and connector actions require approval from the owner or an admin in the chat. It also includes MCP connectors, skills, scheduled tasks, memory, wiki, file storage, and task boards. Agents have their own computers and browsers they can take over, and can also use Claude Code to think on connected computers.
Why it matters Agents are no longer just sidebar bots: they work in channels visible to everyone, and sensitive operations like merges and deployments require human approval.
Stars 1182 (+696 / 24h)
Reasons
- 696 new stars in 24 hours (indicating attention only); the repository was created on 11/9, so reading the README content is normal
- Self-hosting relies only on Postgres; install on a Linux server with one command and connect to a Tailscale internal network
Usage
- Run curl -fsSL https://agentes.chat/install | sh on a Linux server, then enter your own model keys in the app; you can also use the official cloud service directly
Risks
- AGPL-3.0 license: if you modify the code and provide services externally, you must open-source your modifications
- Agent browsers, email, Postgres connectors, mobile and desktop apps are still in beta; Codex integration is also in testing