Skip to main content
Install the Dari CLI, then launch OpenClaw through your organization’s default router:
OpenClaw runs every chat, channel, and scheduled task through one long-lived Gateway that reads ~/.openclaw/openclaw.json, so a single launch cannot route only itself. dari openclaw writes Dari into OpenClaw’s config instead, then starts the installed openclaw executable with every following argument. The write goes through openclaw config patch, which validates it and hot-reloads a running Gateway, so your channels route through Dari too. Your other providers, channels, and settings are kept. Setup adds a dari provider under models.providers that offers your router’s models and makes dari/dari/routing (or your router’s first model, without Autorouting) the primary model in agents.defaults.model. Fallbacks you configured stay in place. The model setup replaced is saved in ~/.openclaw/dari-setup.json. To set up without launching, or to undo it and restore your previous primary model:
On first use, Dari signs you in if needed and creates and privately caches a Routing key, which setup writes as the provider’s apiKey. Gateway includes personal subscriptions, Dari managed billing, and bring-your-own provider keys; the $20/month individual Autorouting plan unlocks the dari/routing router. Gateway serves any model enabled on your router: switch between them with /model, or pick dari/dari/model/<model>. To subscribe, run dari openclaw --connect and choose Get Autorouting (or press a in the setup model picker). On first launch, setup opens an Accounts page where you can connect one or more ChatGPT subscriptions, connect an API key (Vercel AI Gateway, OpenRouter, or direct provider keys), or choose Dari managed billing. Run dari openclaw --connect to open it again. Each launch refreshes the provider, so run dari openclaw or dari openclaw --setup again after changing the router’s models. OpenClaw sends Chat Completions requests through the router at /v1. Setup configures a one-million-token context window and turns on prompt cache keys and session-affinity headers, which group all requests from one OpenClaw session under the same conversation in Activity.

​
Manual Configuration

If you do not want to use the launcher, create a Routing API key and patch it into OpenClaw’s config:
"reasoning": false keeps OpenClaw from sending a fixed reasoning level; Dari still chooses one per request. To use a different model set, edit the default router or Create A Router. To launch dari openclaw on that router, add OpenClaw to its card on the Routers page or run dari router assign-agent <router_id> openclaw; the X on the card moves it back to the default router. With manual configuration instead, point baseUrl at https://routing.dari.dev/{router_id}/v1.