Buzz Review (2026): Pricing, Features & Honest Verdict

What is Buzz?

Buzz is a workspace where your team, your repositories, and your AI agents all live in one place. Block announced it on 21 July 2026, and the pitch is blunt: stop paying Slack for conversation and GitHub for code when the two never talk to each other. Buzz merges them. Channels, direct messages, canvases, git patches, workflow runs, and approvals are all stored as signed events on a single relay.

The part that makes Buzz different from every other Slack clone is the identity model. It runs on Nostr, so every participant gets a cryptographic keypair that belongs to them, not to the platform. That applies to agents too. When Claude Code opens a patch in your repo, the patch is signed by that agent’s own key, and the signature stays attached forever. You get an audit trail that tells you which agent did what, on which branch, with whose approval. We have not seen another tool do this properly. Most competitors bolt agents on as webhook bots with a shared service account, which means six months later nobody can tell you why a file changed.

What Are Buzz’s Key Features?

One identity that travels with you

Every human and every agent in Buzz holds a keypair. Your history, your reputation, and your messages are tied to that key rather than to a workspace admin panel. Leave the company and your identity still works elsewhere. Self-host the relay and you own the whole event log. This is genuinely useful for contractors and open-source contributors who currently juggle four different Slack accounts, and it is the feature most likely to survive if the rest of Buzz gets rewritten.

Agents that work as teammates, not bots

Buzz ships native support for three agent harnesses: Anthropic’s Claude Code, OpenAI’s Codex, and Block’s own Goose framework, which has picked up more than 50,000 GitHub stars since it launched in January 2025. Agents join channels, read the thread history, open patches, and respond to reactions. They are model-agnostic, so you are not locked into one provider. In practice this means you can drop an agent into a bug channel, let it read the last forty messages of context, and have it produce a patch that carries its own signature. Compare that to pasting a stack trace into a chat window and copying the answer back out.

Git that lives inside the conversation

Buzz implements NIP-34, the Nostr standard for git. Patches, reviews, and branch events are signed and stored in the same log as the chat around them. A feature branch effectively becomes a room. When somebody asks why a decision was made in March, you search once and get the message, the patch, the review, and the approval together. That is the actual pitch, and when it works it is a real improvement over hopping between a Slack thread and a pull request that reference each other badly.

YAML workflows triggered by chat

Workflows are defined in YAML and fire on messages, reactions, schedules, or webhooks. A common pattern from the docs: an agent drafts release notes when a tag lands, posts them in the release channel, and waits for a human thumbs-up before publishing. It is less powerful than GitHub Actions today, but it runs where the conversation already is, which removes a context switch that most teams do not notice they are paying for.

Search across conversations, patches, and approvals

Because everything is one event log backed by PostgreSQL full-text search, a single query returns chat messages, code patches, workflow runs, and approval records together. The use case Block leads with is incident memory: search six months back and get the receipts, not just somebody’s recollection in a thread. Anyone who has tried to reconstruct a decision from Slack search and a closed pull request will understand why this is the feature worth caring about. Canvases and media with frame-anchored comments are indexed too, so design feedback does not fall out of the record.

How Much Does Buzz Cost?

Buzz is free. The code is Apache 2.0 licensed at github.com/block/buzz, and you can self-host the entire stack or use Block’s managed version. There is no per-seat charge and no published paid tier.

That is a serious number when you look at what it replaces. Slack Pro runs $8.75 per user per month, Slack Business+ is $15, and GitHub Team is $4 per user per month. A twenty-person engineering team on Slack Pro plus GitHub Team is paying roughly $3,060 a year before anyone adds an AI seat. Buzz costs nothing in licence fees. It does cost you infrastructure and attention: the self-hosted stack needs Docker, Rust 1.88 or newer, Node 24, PostgreSQL for events and search, Redis for pub/sub and presence, and S3 or MinIO for media. Budget an afternoon for the first deploy and someone’s ongoing time for upgrades. “Free” here means free as in puppy.

Who is Buzz Best For?

Use Buzz if you are a small-to-midsize engineering team already running coding agents daily and you are tired of agent work vanishing into an unattributable blob. The signed audit trail pays for itself the first time you have to explain an incident. It also fits teams with a self-hosting habit already, the sort who run their own Postgres and consider that normal rather than a burden.

Use Buzz if you are an open-source project or a distributed collective where portable identity actually matters. Contributors who do not want a corporate Slack account can join with their own key and keep their history.

Use Buzz if you work in a regulated environment where “prove an AI did not touch that” is a question you expect to be asked. Signed events at the message level are a stronger answer than an application log you control and could have edited. We would still want a compliance officer to sign off before anyone relies on that argument, but the raw material is there in a way it is not anywhere else.

Skip Buzz if you need something stable this quarter. Version 0.4.21 is pre-1.0 and behaves like it. Desktop clients exist for macOS, Windows, and Linux, but mobile is unfinished, git hosting is still in development, push notifications are incomplete, and web-of-trust reputation and advanced approval gates are listed as not done. Skip it if your team is mostly non-technical, because sales and marketing will not tolerate a relay that occasionally needs a restart. And skip it if nobody on the team wants to own the infrastructure, because there is no vendor to call at 3am.

Best Buzz Alternatives

Slack

The obvious incumbent. Slack Pro is $8.75 per user per month billed annually, and Slack now bundles AI search and summaries on paid plans. It is more polished than Buzz by an enormous margin, has mobile apps that work, and integrates with everything. What it does not do is treat agents as first-class identities or give you a signed record of what an agent changed. If your priority is reliability over provenance, stay on Slack.

Mattermost

The closest self-hosted comparison. Mattermost is open source, runs on your own hardware, and has a mature free tier plus a Professional plan around $10 per user per month. It is a better chat product than Buzz today and has years of production hardening behind it. It has no native agent identity model and no git integration in the same event log, so you end up wiring those together yourself.

Zulip

Zulip is free and open source with a self-hosted option and a Cloud Standard plan at $8 per user per month. Its threading model is the best in the category, which matters more than people expect once a channel gets busy. Like Mattermost it treats bots as bots. Pick Zulip if the thing you actually hate about Slack is the threading, not the identity model.

Final Verdict: Is Buzz Worth It?

Buzz is the first tool we have looked at that treats the “who did this, and can you prove it” problem as the main event rather than a compliance afterthought. Cryptographic identity for agents is not a gimmick. As soon as three agents and six people are touching the same repo, attribution stops being a nice-to-have. Block has also picked the right integrations: Claude Code, Codex, and Goose cover most of the serious agent usage happening right now, and staying model-agnostic means the choice does not rot in six months.

But it is early, and pretending otherwise would do you a disservice. Version 0.4.21 with unfinished mobile clients and in-progress git hosting is not something you migrate a company onto this quarter. Our recommendation: install it alongside what you already use. Point one repo and one squad at it, run your agents through it for a month, and see whether the audit trail changes how you work. If it does, you will know well before Buzz hits 1.0. If you need a finished product today, bookmark this and check back in six months.