ZCode

Connect the ZCode coding agent to Routero so its model calls run through your gateway — attributed, budgeted, and logged — while ZCode keeps handling repository access and file edits locally.

ZCode connects through a custom OpenAI-compatible provider that talks the Chat Completions API (/chat/completions).

What you must set

In ZCode, open Settings → Model settings → Add provider → OpenAI:

Setting Value
Base URL https://api.routero.dcsmartvision.com/v1
API format Chat completions (/chat/completions)
API key a Routero virtual key
Model list add each model Routero serves, e.g. deepseek/deepseek-v4-flash

Then switch the model in the main dialog to the provider/model pair you just added (e.g. OpenAI/deepseek/deepseek-v4-flash) and send a test prompt.

Only models added to the provider’s model list are selectable. Add every model string your team is approved to use — any model Routero serves works.

Create a key for each developer

In the dashboard, open API Keys and create a virtual key per developer — scoped to their team, limited to the approved models, with an optional budget. Each developer’s ZCode traffic is then attributed individually, and requests appear in Logs routed through Routero rather than straight to the underlying provider.

Gateway policies on ZCode traffic

Policy-bound AI capabilities attached to the key apply to ZCode traffic:

  • Prompt management — injected instructions are honored.
  • Guardrails — enforced per request. Switching a guardrail between Block and Monitor takes effect on subsequent requests without restarting the conversation; blocked requests surface as Request blocked by content filter.
  • Token saving — exact and semantic caching both apply.

→ Calling the API for the base URL and auth model. → Claude Code, Kimi Code, and Codex for the other agents.