1 pointby simonstaton4 hours ago1 comment
  • simonstaton4 hours ago
    I just released v2 of ClaudeSwarm, a self-hosted open-source tool for orchestrating Claude-based agent swarms.

    V2 adds:

    Swarm graph visualization

    Usage + cost tracking

    Built-in billing support

    Would love feedback from anyone building multi-agent systems.

    • verdverm4 hours ago
      I think some questions every project like yours will need to answer...

      1. Why should I choose your 2-day old project over the 1000s of other clawd/cursor operators creating the same thing?

      2. Why should I trust a 2-day old project for anything?

      3. Why should I choose any of 1000s over having my own Ai generate the same thing?

      • simonstatonan hour ago
        Okay I will try to answer this my best, I'm a principal software engineer with 17+ years experience at companies like Google. I've yet to find an agent orchestration harness that operates quite like what I've built here.

        Why is this different to all the other garbage? I've built the message bus from graph theory allowing operators to see the entire topology for optimal orchestration. Also having this boot from a cold start in the cloud makes it accessible from anywhere. I've also built an agent persistence system so that sessions are persisted across server restarts and if the server is shut down all agents will recover.

        The update today has shipped billing so you can see how much each agent is costing along with a graph view to see the entire swarm and better scaling for performance.

        The next release will be horizontal scaling and moving the message bus out of memory to accomplish that. What I can tell you is this 2 day old project was the continuation of 2 months of my work.

        • simonstaton43 minutes ago
          One last thing to add that I missed :) most other tooling runs the SDK directly, I found I was getting way better agentic performance from Claude Code CLI so have developed a number of safe guards and barriers to enable this to run safely