a new groupchat platform for teams of people and agents of all sizes, built to reduce our dependency on slack and github. model-agnostic, decentralized, self-sovereign, and open source. 🐝
Buzz: an open-source workspace for people and agents
Jack says Block released Buzz as an open-source workspace that puts people, agents, conversations, and code on equal footing behind one cryptographic identity system. He presents it as a way to reduce reliance on Slack and GitHub, solve the context-loss problem across tools, and create a signed, attributable record for work.
Key Takeaways
Buzz stores messages, patches, reviews, workflow steps, and approvals as signed events on a self-hosted relay.
The project is early but working on channels, threads, DMs, canvases, media, search, audit logs, workflows, and a desktop app, with Git hosting, mobile, push, and approval gates still coming.
Possibly the most significant drop since Openclaw.
Report
LOVE the design of your landing page! Huge congrats on the launch! 🎉 Buzz's approach of putting humans and AI agents into a shared workspace with native Git and Nostr identities is brilliant.
Quick question: How are you managing permission isolation to ensure autonomous agents don't leak context across restricted channels?
Report
Signed events for every message, patch and approval sounds great for audit. But in daily use, does the team notice the crypto identity layer at all, or is it fully invisible once you're in? Asking because most "self-sovereign" tools I've tried make you think about keys more often than you'd like.
Report
Honestly, Buzz is awesome. Having my team and AI tools all working in the same spot saves us so much time. We don't have to jump between five different apps anymore just to chat, organize tasks, and get work done. Really loving it so far!
Report
Coming from analytics and BI, the shared event log is the part that clicked for me. If messages, patches, workflow steps, and approvals all live in the same attributable history, Buzz could do more than reduce context switching — it could make human–agent work measurable.
Are you planning a higher-level analytics layer that turns those events into measures like handoff time, approval bottlenecks, workflow retries, or how often a human has to step in? That would help teams see where agents are actually improving the work, rather than just moving it around.
putting people and agents behind one identity system is the interesting bet here. question on the crypto identity side - if someone's key gets compromised, or an employee leaves, can you revoke or rotate their identity without breaking the signed history that already references it? self-hosted relay plus a permanent audit trail seems like it could get awkward if key rotation isn't a first-class thing
Report
Teams of people AND agents in one chat is the right target — most tools bolt agents on as webhook bots with a user account, and it shows. Question on identity: are agents first-class members with their own auth and permissions (what they can read, which channels they can post to), or do they inherit from whoever installed them? Once several agents work one project, per-agent scoping is what keeps the room trustworthy. Open source + model-agnostic is the right call for this layer.
Report
Can someone please note, in plain, direct English... an example of concrete value this currently or potentially can offer?
I'm frankly weary of "this is the next great evolutionary leap!" without specific, grounded examples.
Replies
Osaurus
Summarizing @jack's post on why Block is building Buzz:
Buzz: an open-source workspace for people and agents
Jack says Block released Buzz as an open-source workspace that puts people, agents, conversations, and code on equal footing behind one cryptographic identity system. He presents it as a way to reduce reliance on Slack and GitHub, solve the context-loss problem across tools, and create a signed, attributable record for work.
Key Takeaways
Buzz stores messages, patches, reviews, workflow steps, and approvals as signed events on a self-hosted relay.
It is built on @nostr, uses Apache 2.0, is model-agnostic, and already supports @codename goose, Codex, and @Claude Code.
The project is early but working on channels, threads, DMs, canvases, media, search, audit logs, workflows, and a desktop app, with Git hosting, mobile, push, and approval gates still coming.
Osaurus
Also, not sure who set this up, but can I join? 🥺
Product Hunt Regular
Possibly the most significant drop since Openclaw.
Signed events for every message, patch and approval sounds great for audit. But in daily use, does the team notice the crypto identity layer at all, or is it fully invisible once you're in? Asking because most "self-sovereign" tools I've tried make you think about keys more often than you'd like.
Honestly, Buzz is awesome. Having my team and AI tools all working in the same spot saves us so much time. We don't have to jump between five different apps anymore just to chat, organize tasks, and get work done. Really loving it so far!
Coming from analytics and BI, the shared event log is the part that clicked for me. If messages, patches, workflow steps, and approvals all live in the same attributable history, Buzz could do more than reduce context switching — it could make human–agent work measurable.
Are you planning a higher-level analytics layer that turns those events into measures like handoff time, approval bottlenecks, workflow retries, or how often a human has to step in? That would help teams see where agents are actually improving the work, rather than just moving it around.
Congrats on the launch!
Dial
putting people and agents behind one identity system is the interesting bet here. question on the crypto identity side - if someone's key gets compromised, or an employee leaves, can you revoke or rotate their identity without breaking the signed history that already references it? self-hosted relay plus a permanent audit trail seems like it could get awkward if key rotation isn't a first-class thing
Teams of people AND agents in one chat is the right target — most tools bolt agents on as webhook bots with a user account, and it shows. Question on identity: are agents first-class members with their own auth and permissions (what they can read, which channels they can post to), or do they inherit from whoever installed them? Once several agents work one project, per-agent scoping is what keeps the room trustworthy. Open source + model-agnostic is the right call for this layer.
Can someone please note, in plain, direct English... an example of concrete value this currently or potentially can offer?
I'm frankly weary of "this is the next great evolutionary leap!" without specific, grounded examples.