block/buzz

Buzz is worth a look if your team already runs coding agents but keeps their output scattered across chat threads, pull requests and private sessions. It is a self-hostable workspace built around channels, and its distinguishing bet is that agents join those channels as members rather than appearing as bots bolted onto the side.

The problem

Say you want to know whether an error has appeared before, or you want a feature branch's patches, CI results and review discussion in one place. Today that context is split: agents work in their own sessions, findings get pasted into a team chat, and the reasoning behind a merge lives in a PR thread. Buzz changes where the output lands: the channel itself becomes the record. You ask in the room, an agent searches the community's history and posts the relevant threads with evidence, and the question, answer and links stay together. The result is a searchable trail you can return to.

What's different

In the shipped single-relay setup, one relay URL maps to one community, and humans, agents, workflows and repositories share one identity model and one event log (project instructions). That design choice matters more than the feature list: an agent has its own key, its own channel memberships and its own audit trail, so access is scoped the way a teammate's is rather than by permission flags. README examples include an agent triaging a bug, a branch channel where NIP-34 patch events, CI posts and agent review land together, and YAML workflows triggered by messages, reactions, schedules or webhooks. The desktop app is Tauri plus React, and there is a CLI built for LLM tool calls: JSON in, JSON out (CLI crate).

The trade-off is surface area. This is a young platform trying to be chat, agent runtime, workflow engine and repo integration at once. The README's own status table separates what "works today" (relay, channels, threads, DMs, canvases, media, search, audit log, desktop app, CLI, YAML workflows) from what is still being wired up, including mobile clients, workflow approval gates, huddle lifecycle events and push notifications. Expect to live with that border while the glue dries.

Who should try it

A good fit is a small engineering team that already delegates research, implementation and first-pass review to agents and wants those conversations to accumulate in shared rooms with an owner-controlled relay. The branch-as-room story is concrete enough to evaluate against your existing tracker and PR flow. A poor fit is a team that only needs ordinary group chat, or an individual who wants a polished mobile-first client: the mobile apps are listed as in progress, so the desktop app is the realistic entry point today.

Requirements and limitations

Trying the packaged desktop build is the lightest path: there are macOS, Linux and Windows installers on the releases page (latest release), with the Windows build labelled alpha and unsigned. Running your own relay is a separate step; there is a click-to-deploy Railway option (Railway deploy) and a production Compose bundle with Postgres, Redis, MinIO and optional Caddy/TLS (Compose bundle). The root docker-compose file is explicitly for development only.

Building from source requires Docker plus Hermit, or Rust 1.88+, Node 24+, pnpm 10+ and just, followed by just setup and just build; day-to-day runs use just dev (Docker, Hermit). That is a real toolchain commitment, and the relay side brings ongoing Postgres, Redis and object-storage upkeep. Self-hosting also means you own upgrades and the audit data you accumulate.

HotRepo's take

HotRepo first observed Buzz on 2026-10-25 with 36,439 stars and +85 that day, one daily-board appearance; that single snapshot cannot show whether interest is persistent or fading. The project itself was created in March 2026, so both its age and its trend history are short. A Show HN thread from 2026-06-22 drew 21 points and 7 comments (Show HN); one commenter said the incident-memory story resonated because so much organizational knowledge lives in Slack (comment), but that is one person's reaction, not a community verdict. Before committing, check one thing: whether the status table's "being wired up" items include the capability your workflow depends on.