Coordinate with protocol, reason with tokens.
Open-source coordination protocol, prompt injection scanner, and managed agent runtime. Built for teams running AI agents in production.
Your agents don't need conversation history. They read shared state, publish events, and claim resources. 3 primitives, zero token overhead.
name: my-project provider: cerebras/llama-3.3-70b hive: main: acp: spec.acp.yaml agents: - role: developer
$ wgl up # Starts coordination server # Spawns agents from spec # Agents coordinate via protocol
No babysitter LLM. Each agent gets a system prompt + live state injection. They publish events when they finish. Other agents wake up, read state, act.
80-300ms coordination overhead. The protocol is free. Only reasoning costs tokens.
3 primitives. Zero token overhead. Agents share state, publish events, claim resources. YAML spec that's smaller than most config files. Your agents coordinate through protocol, not conversation.
Prompt injection scanner. Drop-in proxy. 29 attack categories detected, zero dependencies. npm install @honeybee-ai/carapace and you're scanning in 5 minutes.
Managed agent runtime. Deploy swarms to the cloud with one command. Durable execution, automatic scaling, real-time telemetry. You write the spec, we run the infrastructure.
wgl is a single binary that does everything. Start local servers, deploy to Colony, scan for injection, publish to the marketplace. If you can wgl up, you're running.
We're onboarding teams one at a time. Tell us a bit about yourself.