Building a Multi-Agent Development Workflow in Buzz

X postSanketOpinion

Sanket (@tinkerersanky) shares his multi-agent configuration in Buzz, distributing software tasks across six specialised agents. He discusses practices for token management, context preservation, and targeted agent prompting.

Key points

  1. Six dedicated agent roles: assistant (terra-high), builder (sol-med), designer (opus-5), planner (opus-5/sol-high), architect (fable-5), and debugger (sol-high).
  2. Targeted scoping with @-mentions: Rather than invoking all agents, users should call specific agents directly with defined task scopes.
  3. Context preservation: Designating a single main communicator or orchestrator helps keep overall context consistent.
  4. Shifting development paradigms: Multi-agent platforms do not fully replace IDEs yet, but represent an evolutionary step toward agent-driven workspaces.
  5. Continuous experimentation: Tooling, token usage workarounds, and inter-agent parallelism remain rapidly evolving and non-final.

Why it matters

Multi-agent systems represent an emerging workflow model where distinct model architectures are assigned to specific development phases based on their comparative strengths.

Analysis

Sanket's practical setup demonstrates how modern workflows map specific models (like opus-5, fable-5, or sol-high) to functional roles such as architecture or debugging. The author's thread emphasizes that context dilution and token consumption remain critical constraints requiring deliberate prompt scoping and orchestration.

Original post

Sanket

@tinkerersanky · 9 August 2026

I am loving Buzz. Made a team of 6 agents - assistant (terra-high) - builder (sol-med) - designer (opus-5) - planner (opus-5/sol-high) - architect (fable-5) - debugger (sol-high) Seems like this may become my default work environment going ahead. Still testing. t.co/U0UPKOJbQt

564 likes52 replies1.3k bookmarks135.9k views

View on X
More posts in the thread (10)
  1. @OzolinsJanis 🙌 wc
  2. @royale_wit_whiz yes it takes a bit of time to get through you, but then suddenly it starts making a lot of sense
  3. @Cyber_Kel_ that sounds amazing!
  4. @BainterSAT @tlongwell_bzz yes, working my way through it by right instructions and finding a flow which doesn't do that. don't call all agents, use @ to just choose who you want to talk to and give defined scope.
  5. @sam_sonivalabs it’s made for that 😄
  6. @oletillmann That's a good setup. Yeah, having a single main communicator / orchestrator is the right way to maintain context. Thanks for the pro tip!
  7. Here's an infographic I made to simplify buzz for myself using ChatGPT image. It also has tips to save token usage. t.co/dqcH0jPxbL
  8. @wesbillman that's a super tip. thanks. will set that up 👍
  9. @AugustoVivaldi yes totally agree with this. It's a new paradigm. It doesn't entirely skips the IDE but is one step towards that. And processes or workarounds for context management, token usage, and parallelism between agents will develop as we go.
  10. @piotryordanov Testing is continous. In AI you are always testing, daily Nothing is stable. Yet
Selected replies (5)
  • @oletillmann @tinkerersanky Love Buzz too. My current agent team: Bessy - Chief of Staff & orchestrator (Fable 5) Rocky - Frontend Design (Kimi K3) Bumble, Fizz & Honey - fast generalists (GPT-5.6 Sol Medium) Sky - large-scale data analysis (DeepSeek 4) Jakob - quality control (GLM 5.2) The interesting
  • @wesbillman @tinkerersanky Awesome! To avoid having to specify the team members and things like "local path" and "git" you can create a Channel Template (Settings->Channel Templates). When you create new channels you can get everything you need each time.
  • @AugustoVivaldi @tinkerersanky 🐝 I’m still a little cautious about using multiple harnesses intensively, especially regarding context management, token usage, and parallelism between agents. But I strongly believe in Buzz’s potential and vision. The project is evolving very quickly, and I’m confident the
  • @royale_wit_whiz @tinkerersanky Once it clicks it’s fantastic
  • @Cyber_Kel_ @tinkerersanky I just installed the buzz relay on my vps. Trying to get an agent to use my local LLMs just for S’s and G’s. Thanks for sharing.

Mentioned links

Links and tools from this entry

Summary generated by AI (Gemini 3.8 Flash, 27 September 2026). It may contain errors – the original source is authoritative.

Related entries

Similar in content, found by the AI search.

  • Video

    Video:Tonbi's AI Garage

    Setting Up Hermes Agent in Buzz for Multi-Agent Collaboration

    A video tutorial walks through integrating Nous Research's Hermes Agent into the Buzz collaborative workspace via the native gateway. Along with solving a common missing CLI prerequisite, it demonstrates multiple agents collaborating and cross-checking research across different machines.

    AI & LLMs· Tutorial

  • Repository:block/buzz

    Buzz: Self-Hostable Nostr-Based Workspace for Humans and AI Agents

    Block has released Buzz, an open-source workspace built in Rust where human developers and AI agents collaborate in the same rooms. Operating as a Nostr relay, Buzz logs every message, workflow step, review approval, and git event as a signed event within a single, unified audit trail.

    32.2kstarsRust

    AI & LLMs· Tool

  • X post:Luna Tech

    OpenAI Workshop: Building Agentic Systems with Codex

    Luna Tech highlights a one-hour OpenAI workshop demonstrating how to build agentic systems with Codex from scratch, covering subagents, prompt patterns, persistent goals, and automated loops.

    4kbookmarks151.7kviews

    AI & LLMs· Resource

Back to the knowledge base

Let's talk about your project

Get in touch

Locations

  • Mattersburg
    Johann Nepomuk Bergerstraße 7/2/14
    7210 Mattersburg, Austria
  • Vienna
    Ungargasse 64-66/3/404
    1030 Wien, Austria

Parts of this content were created with the assistance of AI.