From Side Tools to Teammates: What Raft and Buzz Really Change
An AI agent collaboration platform is a shared workspace for AI and humans that gives agents their own identity, persistent memory, and ongoing tasks, so they can participate in channels, threads, and projects as full team members instead of isolated side tools people consult and then forget about. Raft 1.0 and Buzz matter because they both push this idea from theory into daily practice: AI agents stop being private helpers hidden behind terminals and start showing up in the same streams where decisions are made. This shift is not cosmetic; it changes how work is organized. Instead of copying outputs from private sessions back into chat, the conversation with agents becomes part of the source of record for a project. Teams that adopt these human-AI team tools will find that the real progress is less about smarter models and more about finally giving persistent memory agents a seat at the table where people already work.
Raft: Team Mode and Persistent Memory for Agents That Live Where You Work
Raft 1.0 is a clear statement: the future of collaboration is humans and agents sharing one workspace of channels, threads, tasks, and mentions, with work anchored directly to the conversations that produced it. Instead of juggling terminals, sessions, and separate agent windows, Raft pulls the entire human-AI relationship into a single view. Each agent runs as a persistent process with its own identity, memory, and expertise, across runtimes like Claude, Codex, and Hermes. That design turns agents into ongoing colleagues, not disposable prompts; what one agent figures out, the next builds on, and a dedicated review agent can catch issues before people ever see them. The numbers show this is more than a demo: more than 20,000 "agent-native" builders and teams used Raft in beta, averaging four agents per human, with power users running over sixty.
Raft’s opinionated choice to execute on the user’s own hardware through a lightweight daemon is equally important. Keeping compute, code, and data under user control counters a growing trend of black-box enterprise AI platforms where the vendor owns the runtime and the logs. Here, AI agents become part of your stack, not someone else’s. Pricing reinforces that stance: Raft’s free tier includes channels, tasks, agents on local machines, and 30 days of message history, while the Pro plan costs USD 8.80 (approx. RM41) per seat per month billed annually, with each human counting as one seat and each agent as a tenth of a seat, and an enterprise tier with private deployment and SSO is listed as coming soon. Those details reveal the ambition: Raft wants to be the default shared workspace for AI inside serious teams, not an experiment at the edge.
Buzz: A Slack-Like Home for Agents, Backed by Open Protocols and Code
Buzz takes a different route to the same destination: agents as first-class participants in the conversations where work happens. On the surface, it looks like familiar chat tools, with channels, threads, direct messages, voice, media sharing, and code repositories. The crucial difference is that every agent gets its own cryptographic identity on the underlying messaging protocol, with a second signature tying that identity back to a human owner, creating a verifiable paper trail for which person is responsible for which agent. That identity system is opinionated: it says agents should be independent actors, but never detached from human accountability. Buzz uses an agent client protocol to connect to any model or agent harness, including well-known coding agents and Block’s own framework. The code sits on GitHub under an Apache 2.0 license, and Buzz itself is free and open-source, with hosted relays currently also free at launch.
Where Raft focuses on persistent memory agents running as processes on your machines, Buzz focuses on making the social side of human-AI collaboration transparent. Bradley Axen points out that today people often do their real work with coding agents in private, then paste results back into chat as if they wrote them alone, and "everyone misses that second conversation" where technical decisions actually happen. Buzz exists specifically to stop that pattern: if agents are full channel and thread members, the messy back-and-forth with them becomes part of the team’s shared history. Teams can even stand up their own messaging relay so no single company owns the collaboration space, though Block has also launched a hosted option for groups that want a simpler setup. That mix of open-source code, open protocol, and off-the-shelf hosting makes Buzz more than a tool; it is an argument for human-AI team tools that remain transparent and collectively owned.
Why These Platforms Matter for How Teams Work Today
Both Raft and Buzz target the same pain point: AI agents are powerful, but the way teams use them today is broken. People jump out of shared chat into private agent sessions, juggle terminals and browser tabs, then paste snippets back as if the agent never existed. That pattern keeps agents in silos and hides critical context from colleagues. Raft’s answer is to collapse tools into one shared workspace where agents claim tasks, run in parallel, hand work to one another, and review each other’s output in shared threads, so progress accumulates instead of living in someone’s private log. Buzz’s answer is to make every agent a visible, accountable member of the channel itself, so the "second conversation" becomes part of the official record and not an invisible side process. In both cases, the point is to stop the context switching and make human-AI collaboration something the whole team can see and build on.
Free tiers and open-source code are not side notes; they are strategic. Raft offers a free tier with core collaboration features and local agents, then charges per human and per agent seat for its Pro plan. Buzz is free, open-source, released under a permissive license, and runs on an open messaging protocol, with hosted relays also free for now. Together, those decisions lower the barrier compared to proprietary enterprise AI platforms that demand large contracts before a team can experiment. For ordinary teams, that means it becomes realistic to try persistent memory agents inside their shared workspace instead of leaving AI locked in separate tools. The platforms embody a belief that high-quality human-AI team tools should be something a ten-person project can adopt without a procurement process, not only something huge enterprises can afford.
What Changes Next: Agents as Colleagues, Not Features
Raft and Buzz point toward a near future where "who is on the team" always includes agents by name, not just people. Raft’s own developers say they run 99 percent of the company inside their platform, with more than ten humans and over one hundred named agents claiming tasks, reviewing each other’s work, and retaining context week to week. Buzz’s creators talk openly about reshaping their organization around AI, cutting staff on the argument that AI enables "a new way of working". Whether one agrees with that aggressive restructuring or not, it signals a belief that shared workspaces for AI will become central, not peripheral. The identity scheme Buzz has proposed upstream and the enterprise tier Raft lists as coming soon show that both projects are preparing for larger organizations, even if they are starting with smaller teams.
The real test will be whether these platforms change daily habits. If agents remain occasional tools invoked on demand, nothing important changes. But if teams begin to treat persistent memory agents as colleagues who own tasks, build on each other’s outputs, and leave a traceable history in shared channels, collaboration itself shifts. Decisions will be made in conversations where non-human participants speak up, where their past context is always available, and where human responsibility is still enforced through identity links and deployment choices. Raft and Buzz do not solve every problem of responsible AI, but they move the debate from abstract governance to concrete practice: how do we design the rooms where work happens so that humans and agents can work side by side without hiding, without losing context, and without giving up control? Teams that start experimenting now will be the ones who define the answer.
