Dorsey's Block launches Buzz, AI chat rival
Buzz puts employees and AI agents in one open-source group chat, with agents holding their own identities and permissions, not a shared bot.
Block, Jack Dorsey’s fintech company, released Buzz on July 21: a free, open-source (Apache-2.0) group chat platform built on the decentralized Nostr protocol that puts human teammates and AI agents in the same conversations. Rather than treating agents as chatbots bolted onto a channel, Buzz gives them their own identities with defined permissions, so an agent can post updates, review code, run approved automations, and join threads the same way a person would, with an audit log tracking what it actually did.
The platform is model-agnostic, supporting Claude Code, OpenAI Codex, Block’s own goose agent framework, or any other agent a team wants to plug in. Desktop builds are available now for macOS, Windows, and Linux, with teams able to self-host their own instance or use Block’s managed service through buzz.xyz. Current features include channels, threads, direct messages, shared canvases, search, and YAML-based workflows, positioned as a direct challenge to Slack for messaging and GitHub for code review.
The interesting design choice here isn’t the Slack-versus-GitHub framing, it’s the identity model: an agent with its own scoped permissions and an audit trail is a meaningfully different security posture than letting an AI work in your team chat through a shared bot token that inherits whatever access got set up first and never gets revisited. That’s the same principle behind giving your AI agents their own keys: if you’re already running agents alongside humans in Slack or Discord, the identity-and-permissions model is worth copying even if you never touch Buzz itself.