Five runtimes, run in parallel and matched to the work.
Standardizing on one runtime is a single point of failure. We operate five, route each task to the runtime that fits, and stay resilient when any one underperforms.
OpenClaw
Primary workhorseDrives most conversational flows and tool use across client deployments. The default runtime for end-to-end agent work.
OpenCode
Code & automationHandles code generation, refactors, and developer-facing automation where correctness and structure matter.
ClawCode
Hybrid runtimeBridges conversational and coding workloads, used when a task needs both natural dialogue and real code output.
Hermes
Messaging backbonePowers async messaging pipelines and webhook-driven integrations across every channel we ship to.
Pi
Lightweight agentFast, focused agent for narrow, high-frequency tasks where latency and cost are the priority.
Why multiple runtimes?
Each runtime has strengths — conversation, code, latency. We match the runtime to the task instead of forcing one model everywhere.
If a runtime degrades or rate-limits, work fails over to another. No single dependency takes the whole system down.
Lightweight runtimes like Pi handle high-frequency work cheaply; heavier runtimes are reserved for tasks that need them.
Put a company of agents to work.
Tell us the workflow you want run autonomously. We'll scope the agents, runtimes, and governance — and operate it for you.