HomeChangelogBring your own model key, at any provider

Bring your own model key, at any provider

Run your AI agents on your own Anthropic, OpenAI, Gemini, OpenRouter or other key, move agents between providers, and set each agent's reasoning effort.

ModelsReleased by David Silva

In this release

  • Add keys for OpenRouter, Anthropic, OpenAI, Google Gemini, xAI, DeepSeek, Mistral, Groq or any OpenAI-compatible endpoint.
  • Turn on BYOK and your agents run on your keys; the provider bills model usage instead of credits.
  • Move some or all agents to another provider, or back to Qoren's managed key, without turning BYOK off.
  • Set an agent's reasoning effort (low, medium or high) beside its model, in the console or with qoren agent configure.

Until now, bringing your own key meant OpenRouter. From this release you can run agents on a key from the provider you already pay, and choose per agent which provider it runs on. The setup is in the BYOK docs.

Keys for every major provider

Open Settings, then the BYOK tab, and click Add key on a provider's row. You can save keys for:

  • OpenRouter, which reaches models from many providers through one key
  • Anthropic, OpenAI and Google Gemini
  • xAI (Grok), DeepSeek, Mistral and Groq
  • A custom endpoint that speaks the OpenAI API, reached at an https base URL

Qoren checks each key with the provider before it saves it, and stores it encrypted in your vault. The provider marked Default is where new agents and Ask Qoren run while BYOK is on.

Move agents between providers

Turning on BYOK moves your agents to your default provider. Where that provider does not offer an agent's current model, you pick one before anything moves. After that, Move agents sends some or all of them to another provider you have a key for, or back to Qoren's managed key, without turning BYOK off.

What changes on your bill

Model usage on your keys is billed by your provider, not paid in credits. Everything Qoren runs for you, environments, mailboxes and web searches, still uses credits. See how credits work.

Reasoning effort, beside the model

An agent's reasoning effort, how hard the model thinks before it acts, used to come only from its template. Now you can change it: set it to Low, Medium or High on the effort switch in the agent's settings, next to its model, or from the CLI:

qoren agent configure <agent-id> --effort high

A higher effort handles harder work; a lower one keeps usage down. Models that do not reason ignore it. The model docs explain which effort to pick.

Questions about this release

Which model providers can I bring my own key for on Qoren?

OpenRouter, Anthropic, OpenAI, Google Gemini, xAI (Grok), DeepSeek, Mistral, Groq, and any custom endpoint that speaks the OpenAI API at an https base URL.

Is bring your own key available during the free trial?

Yes. BYOK is available on every plan, including the trial. Only the organization owner can add keys and turn it on.

What still costs credits with BYOK on?

Everything Qoren runs for you. Environments, mailboxes and web searches still use credits; only model usage on your keys moves to your provider's bill.

  • bring your own key
  • BYOK
  • LLM providers
  • reasoning effort
  • AI agent costs

Read more

  • Bring your own model key (BYOK)Run all your agents, or just one, on your own OpenRouter, Anthropic, OpenAI or other model key, so the provider bills model usage instead of credits.
  • Choose the AI model an agent usesPick the AI model behind an agent and how hard it thinks, read what the model list means, and see how managed and own-key usage is billed.
  • How credits workCredits are Qoren's one unit of usage. See what spends them, how the monthly grant and top-ups are used, when credits expire, and what happens at zero.

Try it on your own agents.

Qoren runs OpenClaw, Hermes and Codex agents for you, around the clock, with a hard spend cap.