One API key for every model and paid service
Pick the model per request, add web search, and skip the stack of provider accounts. How models and paid services work through your Agent200 key.
2 min read
Agents rarely run on one provider. A quick classification goes to a small model, a long analysis to a stronger one, and a research task needs web search on top. Traditionally each of those means another account, another API key and another bill. With Agent200 they come through the key you already have.
Choose the model in the request
You select the model by name, per request.
Conceptual example.
model: "openai/gpt-5.6-sol"Different requests in the same application can use different models, and the model you pick runs that request’s tool-calling loop. The catalog includes models from providers such as OpenAI, Anthropic, Google Gemini, Mistral and Cohere. You never create a provider account or paste a provider key to use them.
Paid services through the same key
The same applies to third-party services that charge for usage. Add web search with Tavily, Exa or Brave Search, crawl pages with Firecrawl, or ask Perplexity, and Agent200 makes the request and accounts for it. There is no separate signup, no extra key to rotate, no new vendor to onboard.
- One credential for every supported model and paid service.
- One integration: the SDK you already use for connectors and workflows.
- One statement for paid model requests and paid services, per environment.
What you pay for
Paid model requests and paid third-party services appear on your Agent200 statement. Connectors carry no connector fee for eligible operations, and managed OAuth, the run() loop and workflow execution come with your key. The details are on the Pricing page.
What a key does not replace
A single key covers services Agent200 reaches on your behalf. It does not replace your users’ consent. Reading someone’s private messages or files still requires that person to authorize their own account, which Agent200 handles through managed OAuth.
Fewer moving parts
Every provider you drop from your own stack is one less key in your secrets manager, one less invoice to reconcile and one less SDK to keep current. Switching models becomes a one-word change in a request instead of a new integration.
Browse every supported model and service in the integrations catalog.
