AMD Agent Studio

Watch the agent work

Every model call, tool, subagent and routing decision.

Idle

00:00

Subagents appear when OpenClaw delegates work.

Ready for a live task

OpenClaw calls one endpoint. The semantic router chooses local Qwen, Fireworks GLM, or premium Kimi—and reports why.

Routing intelligence

Why each call went where it did—and at what effort.

MODEL CALLS 0

CLOUD API COST* $0.00000

ALL-KIMI EST.* $0.00000

EST. SAVINGS —

PROMPT 0 OUTPUT 0 REASONING 0

Routed cloud API All-Kimi estimate

Cloud token/API estimates update as model calls complete.

Cost scope

Displayed cost is cloud token/API cost only and excludes local AMD Instinct MI300X hardware, hosting, power, and operating costs. “All-Kimi” reprices the observed tokens at Kimi K3 API rates; it is an estimate, not a second run.

CALL DISTRIBUTION

LATEST ROUTE

Run a task to see the selected model, matched signals, and router-requested reasoning effort.

Watch the agent work

Ready for a live task

Task complete