A new Claude Code plugin from Functio-AI is making some aggressive performance claims. The project, dubbed 'claude-go-brr' and now at version v0.1.1, promises speedups ranging from 2× to 8× across different workflow types—including deep-research tasks, ultracode execution, and main chat agent swarms. The plugin is available via the Claude Code marketplace or direct shell installation.

The Performance Claims Breakdown

The most concrete benchmark comes from a 128-agent computer-use test, where Functio-AI reports an average prompt-to-completion latency speed-up of 5.5×. For Kubernetes-based orchestration, they tested 32 parallel agents building ray-tracing applications to calculate light and shadow maps for home layouts—achieving what they describe as a 2.4× reduction from their baseline (dropping from 680 seconds average latency down to 285 seconds). The plugin also claims 2–5× improvements on main chat agent swarms.

Getting It Running

Installation is straightforward via two methods. Claude Code users can add it through the plugin marketplace with four commands: /plugin marketplace add Functio-AI/claude-go-brr, then /plugin install claude-go-brr@claude-go-brr, followed by /reload-plugins and /claude-go-brr:install v0.1.1. The alternative is a curl-based one-liner pulling the install script directly from GitHub.

Read the Fine Print

Here's where you need to channel your inner security researcher before deploying this anywhere near production. These performance numbers are entirely self-reported by Functio-AI—there's no independent benchmark verification, no methodology documentation detailing hardware specs or baseline Claude Code versions for comparison. The benchmarks describe scenarios (computer-use agents, Kubernetes orchestration) but omit the critical details needed to reproduce them.

Key Takeaways

  • Plugin targets three main areas: deep-research (2–8×), ultracode workflows (2–8×), and agent swarms (2–5×)
  • Hardest data point: 128-agent run showing 5.5× latency reduction in computer-use tasks
  • Kubernetes test with 32 parallel agents dropped from 680s to 285s—a 2.4× improvement
  • Installation is simple via marketplace or shell script for v0.1.1

The Bottom Line

If the numbers hold up, this could genuinely change how teams run heavy agent workloads—but self-reported benchmarks without independent verification are a red flag. Pull the repo, read the code, and run your own tests before trusting it with anything that matters.