Claude Code Provider Switcher
The scenario: you want to run Claude Code, but through a provider and key you chose — DeepSeek, Kimi, GLM, OpenRouter, or a self hosted compatible endpoint — without editing environment variables before every session...
The scenario: you want to run Claude Code, but through a provider and key you chose — DeepSeek, Kimi, GLM, OpenRouter, or a self-hosted compatible endpoint — without editing environment variables before every session and without an official coding-plan subscription. VibeAround's provider profiles and API Bridge make that routing a named, visible choice at launch time.
Claude and Claude Code are Anthropic products. VibeAround is independent software that coordinates local workflows.
The Walkthrough
-
Create a provider profile. In the desktop app's model profile screen:
- Pick a provider from the catalog (Kimi/Moonshot, DeepSeek, OpenRouter, Z.AI/GLM, DashScope, and more — endpoints and models prefilled), or custom for any compatible endpoint.
- Pick the endpoint variant where the provider has several (global vs CN, pay-as-you-go vs coding plan) — the key must match the plan.
- Paste the API key. It is stored locally and only ever sent to that provider by the daemon.
- Choose which models the profile exposes and which is the default. For agents that validate model names, define alias model ids that map to the real upstream model.
Full details in the model profiles guide;
va profileslists profiles from the CLI. -
Launch Claude Code with the profile. Desktop Launch screen: pick Claude Code + workspace + the new profile — or from the CLI:
va launch --profile <name> # launch with the saved launch profile va launch --profile <name> --dry-run # print the rendered plan without launchingThe rendered config points Claude Code's model traffic at the local bridge (
127.0.0.1:12358); the daemon translates to the provider. Keep the daemon running for the session's lifetime. -
Test with a small task before relying on the route. If the agent says "model not found", use the profile's alias model id, not the upstream id.
-
Switch mid-conversation when hosted. In an IM or web chat thread,
/profile --listshows the profiles and/profile --switch <id>re-binds the thread — useful for comparing providers on the same task.
When Not To Switch
Do not add a profile layer when native Claude Code configuration (your Anthropic login) is already the simplest, most reliable choice. Provider switching is most valuable when repeatability, explicit routing, model comparison, third-party keys, or bridge translation matter.
The Same Switch, Everywhere
Profiles are not Claude Code specific. The same profile can launch Codex CLI, Gemini CLI, OpenCode, and the other supported agents, and any OpenAI-compatible tool can point at the profile's local endpoint (shown in the profile UI) to share the same provider setup. Agent, workspace, session, and profile are all selected per launch — visible, reversible, repeatable.
Related Docs
Last verified: v0.7.11
OpenCode Remote Access
The scenario: OpenCode is running next to your repository and tools on the host machine, and you want to check on it from a phone during a break — or hand the session over entirely and continue from the sofa. VibeArou...
Agent launch guide
Agent Launch opens a coding agent CLI in your own terminal — credentials materialized, model routing configured, project integrations installed — instead of hosting it inside the daemon. Use it when you want the full...