Buzz: An Open-Protocol Agentic Workspace
Greg Isenberggo watch the original →
the gist
Buzz is an open-source, agent-native chat application that treats AI agents as first-class team members, allowing for persistent context, swappable model harnesses, and end-to-end software development within a shared workspace.
The Shift from Integrations to First-Class Agents
Buzz distinguishes itself from traditional communication tools like Slack by treating AI agents as core team members rather than external integrations. In this model, agents exist within the same persistent chat context as human users, allowing for seamless collaboration. Users can assign specific system prompts to agents, effectively defining their roles and capabilities within the team structure.
Swappable Harnesses and Context Persistence
A primary technical advantage of Buzz is its "harness" architecture. Users can swap underlying models—such as Claude Code, Codex, or Goose—without losing the context of their ongoing projects or conversations. This addresses the common issue of "model fatigue," where users feel locked into a specific ecosystem. Because the context remains tethered to the project rather than the model, developers can iterate across different engines as new, more capable models emerge.
End-to-End Development Workflows
Buzz integrates Git and project management directly into the chat interface. Agents are capable of creating feature branches, working in parallel worktrees, and pushing code to remote repositories hosted on a user's relay. Vinny demonstrates this by having an agent build a full-stack CRM using the Wasp framework and deploying it to Railway, all initiated via natural language commands within the chat. This capability turns the chat interface into a functional development environment.
Open Protocols and Data Ownership
The platform is built on open protocols, which simplifies the integration of external data sources. Unlike proprietary platforms that require complex API token management and app registration, Buzz allows users to pipe data from external APIs directly into the agent's context. This creates a feedback loop where agents can analyze live data—such as social media performance metrics—and suggest strategic adjustments, effectively automating the "boring" but essential operational tasks of a business.
Collaborative Huddles and Future Work
Buzz includes audio huddles where agents can participate in real-time voice conversations, maintaining access to the full chat history. This feature aims to bridge the gap between static text-based prompting and dynamic, creative brainstorming. While still in early preview, the tool is positioned as a powerful environment for solopreneurs and small teams who need to move quickly and maintain high-fidelity context across their development and operational workflows.