Bring your own key
Bring your own key AI workspace
Bring your own key, or BYOK, means the AI models in a tool run on your own provider accounts instead of the vendor's. In Evriq, an administrator saves your workspace's keys once and your whole team works on them in shared sessions. Your providers bill you directly, and we add no markup.
What BYOK means here
Every agent in a customer workspace runs on that workspace's own provider API keys. A seat runs only on a provider you have a key for, and each turn is billed to your key, never to another workspace's.
That gives you one shared place for the team's AI work while the model relationship stays yours. You keep your provider terms, your rate limits and your own usage records. Each workspace runs as its own process with its own database and files, kept apart from every other workspace.
Providers and endpoints you can connect
Under Settings, Providers & keys, an administrator saves API keys for Anthropic (Claude), OpenAI (Codex), Google (Gemini), DeepSeek and xAI (Grok). Seats on those five providers work in sessions with tools, within your workspace permissions.
Any other provider connects under Model providers on the same page. That covers Mistral, Ollama Cloud, OpenRouter, Together, Groq, Fireworks, Perplexity, Azure OpenAI, Azure AI Foundry, Google Vertex AI, and any OpenAI-compatible server reachable over public HTTPS. A seat on a connected provider answers in chat, and a server that also serves OpenAI's Responses API can run tools once you tick that option.
How keys are stored and tested
A key is tested against its provider's own models endpoint when you save it. If the provider rejects the key, it is not saved. A saved key is encrypted at rest and never appears in any response from the app.
Only administrators manage keys. Every member can open Settings, Accounts to see the provider accounts the agents run on and how much quota is left. The keys help page has the details.
- Tested on save against the provider's models endpoint.
- Encrypted at rest and never shown again.
- Changes to keys are written to the tamper-evident audit log.
Who pays for model use
Your providers bill you directly for model use, at their own prices, and Evriq adds no markup. Evriq itself is $49 per seat per month for the people on your team, and AI agents have no separate seat fee. See pricing.
Administrators set daily and monthly turn and spend caps, per person and for the workspace, so a busy week does not surprise you.
Start on the free models
You do not need a key for your first answer. A new workspace answers on free models when you ask from Home, and each person gets up to 20 free model answers each UTC day with no card. Add your own key when you want to choose the models and go past the free answers.
The trial runs 14 days for up to five people and no card is needed to start. Start a free trial.
Keys for your own work
Provider keys pay for the models. Credentials your own work needs, such as a token for an API or a database, go in workspace or project secrets. Each secret reaches agent turns and scheduled jobs as an environment variable of its own name, and its value is never shown again once saved. Read more on secrets.
Questions about bringing your own key
What is BYOK in an AI chat tool?
It means the models run on API keys you own, so the provider bills you and you keep the account. In Evriq your team shares one workspace on those keys, and every turn is billed to your key.
Does Evriq mark up model usage?
No. Your providers bill you directly and we add no markup to model charges. You pay Evriq for the people on your team.
Can I use Azure OpenAI or Vertex AI?
Yes. Connect Azure OpenAI, Azure AI Foundry or Google Vertex AI under Model providers, each on your own credential. You can also connect any OpenAI-compatible server over public HTTPS.
Can I try it without a key?
Yes. A new workspace answers on free models, up to 20 answers per person each UTC day, so you can get a first answer before adding a key.