Every Clark model and agent, behind one API.
Point any OpenAI-compatible client at the Clark gateway and use Clark Agent or any provider model. Your app keeps the SDK it already has; Clark handles the rest.
One gateway, two ways to run.
Choose who runs the tools: hand the whole job to Clark Agent, or keep the loop in your app and use Clark as a model gateway.
-
Clark Agent as an API
Set model to "clark" or "clark_max" and Clark owns the loop: planning, durable memory, built-in search and browsing, a managed computer, parallel specialist agents, and verified artifacts.
-
Model gateway
Set model to "provider/model" and your app runs the tool calls while Clark handles model routing and retries across frontier providers.
-
OpenAI-compatible
Chat completions and responses work with the OpenAI client you already use. Change the base URL and API key; keep your code.
-
Files, memories, and code
API resources cover models, responses, chat completions, files, memories, and code repositories.
-
Official SDKs
TypeScript, JavaScript, Python, Rust, and Zig clients, plus a copyable integration brief for coding agents.
-
Per-token pricing
Pay per token on every gateway model. The full catalog, cache pricing, and live regional rates are in the app.
Flagship models, per 1M tokens.
| Model | Input · $ / 1M tokens | Output · $ / 1M tokens |
|---|---|---|
| Kimi K3 | $3.60 | $18 |
| GPT-5.6 Sol | $6 | $36 |
| Claude Opus 4.8 | $6 | $30 |
| GPT-5.6 Terra | $3 | $18 |
| Claude Sonnet 5 | $2.40 | $12 |
| Gemini 3.5 Flash | $1.80 | $10.80 |
| GLM 5.2 | $1.17 | $3.67 |
| Qwen 3.7 Plus | $0.38 | $1.54 |
| Qwen 3.5 Flash | $0.078 | $0.31 |
Prices are what you pay through the Clark gateway. Sign in for the full model catalog, cache pricing, and live rates.
Get started
Create an API key and make your first call.