Claude Code
Model-specific setup
and request instructions.
One gateway.
Built for your stack.
Connect your applications and coding agents to AI models through one OpenAI-compatible API. Your next idea starts here.
One endpointExact model IDsYour existing SDK
Your stack, connected.
Quickstartimport OpenAI from "openai"; const client = new OpenAI({ apiKey: process.env.FOUNDRY_API_KEY, baseURL: "https://api.fdy.sh/v1",}); const response = await client.chat.completions.create({ model: "moonshotai/kimi-k2.6", messages: [{ role: "user", content: "Hello, Foundry." }],});https://api.fdy.sh/v1One request shape. Your choice of model.Connected to the models
you build with.
RelaceOpenRouterDeepInfraSiliconFlowNebius Token FactoryAvian
The model directory
More choice.
One place to explore.
Find your next model with current provider routes and published token prices, all in the same directory.
Browse all models4,862 models in the catalog839 routeable · 88 providers
| Model | Type | Input / 1M tokens | Output / 1M tokens | Provider routes |
|---|---|---|---|---|
| moonshotai/kimi-k2.6Moonshot AI | chat | $0.4344-$1.25 / 1M | $2.45-$5.3 / 1M | 13 routeable |
| deepseek-ai/deepseek-v4-proDeepSeek | chat | $1.3-$1.75 / 1M | $2.6-$3.96 / 1M | 11 routeable |
| deepseek-ai/deepseek-v4-pro-0813DeepSeek | chat | $0.243-$1.32 / 1M | $1.683-$4.2 / 1M | 10 routeable |
| openai/gpt-oss-120bOpenAI | chat | $0.037-$0.80 / 1M | $0.17-$0.80 / 1M | 10 routeable |
| deepseek-ai/deepseek-v4-flash-0731DeepSeek | chat | — | — | 1 routeable |
| moonshotai/kimi-k3Moonshot AI | chat | — | — | 1 routeable |
Showing 6 models. Unpublished prices appear as —.Explore the directory
Provider-aware routing
Find your model.
Understand your route.
Compare current routes, published prices, and observed latency. Keep your chosen model identity throughout the request.
| Model | Available provider routes |
|---|---|
| moonshotai/kimi-k2.6 | 13 |
| deepseek-ai/deepseek-v4-pro | 11 |
| deepseek-ai/deepseek-v4-pro-0813 | 10 |
| openai/gpt-oss-120b | 10 |
Route counts share a zero-based scale. Showing the first 4 models in this preview. Missing prices and telemetry remain unpublished.
Keep the model you chose.
Provider selection preserves the exact canonical model ID in your request.
See the evidence behind a route.
Availability, prices, and observed health stay tied to the current registry.
Handle unavailable routes explicitly.
No eligible provider means a clear error your application can act on.
Made for the way you work
Your favorite agents.
Ready for Foundry.
Give your coding agent a model-specific skill with the endpoint, environment variables, and request contract it needs.
Open the integration generatorSKILL.mdCodex
Model-specific setup
and request instructions.
Hermes
Model-specific setup
and request instructions.
OpenCode
Model-specific setup
and request instructions.
OpenClaw
Model-specific setup
and request instructions.
GrokBot
Model-specific setup
and request instructions.
T3Codes
Model-specific setup
and request instructions.
The complete picture
Build with confidence.
Operate with clarity.
Keep the request simple and the details within reach. Keys, usage, and billing belong in the same workflow.
Your keys. Your control.
Create, rotate, and revoke tenant-scoped API keys from one console.
Manage API keysKnow what every request costs.
Inspect token usage, billing lanes, and receipts alongside the work that ran.
Explore usageChoose how you run.
Keep Gateway PAYGO, Foundry Pro, and BYOK billing lanes explicit.
Compare plansWhat is Foundry?
Foundry is an AI gateway with an OpenAI-compatible API at https://api.fdy.sh/v1. Connect your application or coding agent, choose an exact model ID from the current catalog, and send requests through one endpoint.
Can I keep my existing SDK?
Yes. Use an OpenAI-compatible client, set the Foundry base URL and API key, and keep your message payload. The integration generator provides model-specific setup directions for supported coding agents.
Where do model availability and prices come from?
The model directory reads the current registry. Provider routes, token prices, and observed health appear when published. If the current snapshot is unavailable, Foundry leaves the data unavailable rather than substituting a cached catalog.
Does Foundry switch my model if a route is unavailable?
Your request keeps the exact model identity you selected. If no eligible provider route is available for the selected billing lane, the gateway returns an explicit error so your application can decide what to do next.
How does billing work?
Gateway PAYGO and the Foundry Pro coding subscription are separate billing lanes. Gateway access can use usage-based credits, x402 per-request payment, BYOK, or an enterprise commitment. See Plans for lane-specific terms and the model directory for published token prices.
Your next build starts here.
Pick a model. Get a key. Make something new.