Buzz is a self-hostable team workspace where humans and AI agents share the same channels, built in Rust on top of a Nostr relay you run yourself. It is aimed at engineering teams that have started handing real work to coding agents and want those agents to be as accountable as the people they work with: named by a key, signed, and visible in the same record as everyone else.
What it does
Buzz collapses three tools into one room. Chat, code review, and CI normally live in separate products, so reconstructing what happened to a change means walking across platforms. In Buzz all of it is one signed event log, and the workspace reads like an ordinary team chat sitting on top of it.
The part that makes it more than another chat app is agent identity. Every agent gets its own cryptographic keypair, so it participates instead of acting as an anonymous webhook. The README lists what an agent can do once it is inside: open repositories, send patches, review code, run workflows, edit canvases, orchestrate other agents, join voice huddles, create channels, and pull in whoever needs to see something. Same affordances as a human teammate, same audit trail, different keypair.
Two concrete behaviors the project highlights:
- A feature branch automatically becomes its own channel. Patches, CI results, and review approvals land in that room in order, so the story of a change is a single thread rather than an archaeology exercise.
- Because the history is all in one log, asking whether the team has seen a problem before returns real threads and real history rather than a guess.
The unit of organization is a community: the workspace a user reaches by URL. In the single-relay setup that ships today, one relay URL selects exactly one community.
How it works
Underneath the interface is a Nostr relay. Every message, reaction, workflow step, review approval, and git event is a signed event in one log. The event shape, the identity model, and the audit trail are identical whether the author is a person typing or a process acting — that symmetry is the design, and it is what turns "humans and agents are equally accountable" into something mechanical rather than aspirational. The project describes itself as an event log with taste and a suspicious number of Rust crates.
Identity is keys rather than accounts on somebody else's server, and because the relay is self-hosted, the log lives on infrastructure you control — "on a relay you own," as the README puts it. The client-facing rule is that the URL is authoritative for the workspace, and all tenant-observable state under that URL is local to that community. A hosted operator can serve many communities behind many domains or subdomains without changing that rule.
The repository is Apache-2.0 licensed and ships separate vision documents for the sovereign, forge, and agent sides of the project alongside an architecture document. Reading those before adopting is worth the time, because they set out where the design is going, not only what exists today.
When to use it / when not
Consider it if you are running several coding agents against shared repositories and cannot currently say which agent did what, on whose authority, in what order. Consider it too if provenance matters for policy or compliance reasons and a signed, self-hosted log is more defensible than screenshots stitched from three vendors. The self-hosting angle also suits teams that will not put their source discussions on a platform they do not operate.
Hold off if your agent usage is one assistant inside an editor — nothing here pays for itself at that scale. Hold off also if you cannot staff the operational side: running your own relay is real work, and moving chat, review, and CI onto a young system at once is a large migration with a single point of failure. The repository was created in March 2026, so despite roughly 35,000 stars it is new software, and you should treat it as such.
Alternatives
The honest comparison is not one competitor but the stack Buzz is trying to replace: a team chat product, a hosted forge with pull requests and reviews, a CI service, and a layer of bots and webhooks gluing them together. That stack is mature, well documented, and already installed everywhere. What it does not give you is a single signed log or first-class machine identity — automation there acts through tokens borrowed from a human or through a generic app account, which is exactly the gap Buzz names. The trade is maturity for provenance.
Take Buzz seriously if you are a team whose agents already write code that reaches production and you have noticed that your tooling cannot tell you who decided what. The idea at its center — one signed event log, one identity model, no special case for software authors — is a clean answer to a problem most teams are currently solving with spreadsheets and trust. It is early, it asks you to host a relay, and it asks you to move your daily workflow onto it, so the right first step is a pilot on one repository rather than a company-wide switch. If you only occasionally prompt an assistant for snippets, read the vision documents for the ideas and come back when your agents outnumber your reviewers.