Use OpenClaw with WebikAI

OpenClaw takes custom OpenAI-compatible model providers, so its agents can run on WebikAI routes: give the agent a client-labeled key with a spend cap and every call it makes is governed, receipted, and recorded per client.

Setup

Create an account, confirm your email, complete the four-question onboarding, and copy your wk_live_… key (shown once).
Add a supported provider credential for BYOK, fund the managed wallet, or do both for hybrid.
Add WebikAI to OpenClaw's models.providers config:

json

{
  "models": {
    "mode": "merge",
    "providers": {
      "webik": {
        "baseUrl": "https://gateway.webik.ai/v1",
        "apiKey": "wk_live_...",
        "api": "openai-completions",
        "models": [
          { "id": "auto/chat", "name": "WebikAI auto/chat" },
          { "id": "auto/code", "name": "WebikAI auto/code" }
        ]
      }
    }
  },
  "agents": { "defaults": { "models": ["webik/auto/chat", "webik/auto/code"] } }
}

Defining the provider is not enough on its own — OpenClaw only serves models you also allowlist in agents.defaults.models (the webik/… entries above). See OpenClaw's model-providers reference for the full schema.

Models to start with

Alias	Routes to
auto/chat	Balanced general-purpose chat
auto/code	WebikAI lane ranked by coding evidence under your rules
auto/cheap	Aggressively minimize cost
auto/vision	Multimodal — image or document input
z-ai/glm-5.2	Pin an exact model (OpenRouter-style slug)

Honest limit

A persistent agent can spend around the clock, and a spend cap holds estimates — the provider's final reported usage can exceed one. Give an always-on agent its own key with a monthly cap sized for unattended operation, not your whole wallet.

Route receipts and spend controls

Every successful routed response carries the five x-webik-* receipt headers, and per-key or client spend caps can stop a request before it is forwarded. Label the key by client to separate usage in reports.

Full reference: docs · live scores: benchmarks · start free