feat(ai,coding-agent): add Requesty as a built-in provider

packages/ai/README.md

@@ -75,6 +75,7 @@ Unified LLM API with provider collections, automatic auth resolution, token and
 - **xAI**
 - **OpenRouter**
 - **Vercel AI Gateway**
+- **Requesty**
 - **ZAI Coding Plan (Global)** (with separate China provider)
 - **MiniMax** (with separate China provider)
 - **Together AI**
@@ -433,6 +434,7 @@ Built-in providers resolve these env vars (Node.js; in browsers pass `apiKey` ex
 | Baseten | `BASETEN_API_KEY` |
 | OpenRouter | `OPENROUTER_API_KEY` |
 | Vercel AI Gateway | `AI_GATEWAY_API_KEY` |
+| Requesty | `REQUESTY_API_KEY` |
 | ZAI Coding Plan (Global) | `ZAI_API_KEY` |
 | ZAI Coding Plan (China) | `ZAI_CODING_CN_API_KEY` |
 | MiniMax (Global) | `MINIMAX_API_KEY` |

packages/ai/scripts/generate-models.ts

@@ -164,6 +164,23 @@ interface AiGatewayModel {
 	};
 }
 
+interface RequestyModelListItem {
+	id: string;
+	api?: string;
+	input_price?: number;
+	output_price?: number;
+	caching_price?: number;
+	cached_price?: number;
+	context_window?: number;
+	max_output_tokens?: number;
+	supports_tool_calling?: boolean;
+	supports_vision?: boolean;
+	supports_reasoning?: boolean;
+	supports_role_developer?: boolean;
+}
+
+const REQUESTY_ROUTER_URL = "https://router.requesty.ai";
+
 const COPILOT_STATIC_HEADERS = {
 	"User-Agent": "GitHubCopilotChat/0.35.0",
 	"Editor-Version": "vscode/1.107.0",
@@ -1327,6 +1344,70 @@ async function fetchAiGatewayModels(): Promise<Model<any>[]> {
 	}
 }
 
+async function fetchRequestyModels(): Promise<Model<any>[]> {
+	try {
+		console.log("Fetching models from Requesty API...");
+		const response = await fetch(`${REQUESTY_ROUTER_URL}/v1/models`);
+		if (!response.ok) throw new Error(`Requesty API returned ${response.status}`);
+		const data = (await response.json()) as { data?: RequestyModelListItem[] };
+		const models: Model<any>[] = [];
+
+		for (const model of data.data ?? []) {
+			if (model.api !== "chat" || !model.supports_tool_calling) continue;
+
+			const input: ("text" | "image")[] = model.supports_vision ? ["text", "image"] : ["text"];
+			// Requesty prices are $/token.
+			const cost = {
+				input: roundCost((model.input_price ?? 0) * 1_000_000),
+				output: roundCost((model.output_price ?? 0) * 1_000_000),
+				cacheRead: roundCost((model.cached_price ?? 0) * 1_000_000),
+				cacheWrite: roundCost((model.caching_price ?? 0) * 1_000_000),
+			};
+			const contextWindow = model.context_window || 4096;
+			const maxTokens = model.max_output_tokens || 4096;
+
+			// Anthropic models go through Requesty's native /v1/messages endpoint so thinking blocks,
+			// cache_control, and signatures pass through unchanged. Everything else uses chat completions.
+			if (model.id.startsWith("anthropic/")) {
+				models.push({
+					id: model.id,
+					name: model.id,
+					api: "anthropic-messages",
+					baseUrl: REQUESTY_ROUTER_URL,
+					provider: "requesty",
+					reasoning: model.supports_reasoning === true,
+					input,
+					cost,
+					contextWindow,
+					maxTokens,
+				});
+				continue;
+			}
+
+			models.push({
+				id: model.id,
+				name: model.id,
+				api: "openai-completions",
+				baseUrl: `${REQUESTY_ROUTER_URL}/v1`,
+				provider: "requesty",
+				reasoning: model.supports_reasoning === true,
+				input,
+				cost,
+				contextWindow,
+				maxTokens,
+				...(model.supports_role_developer ? {} : { compat: { supportsDeveloperRole: false } }),
+			});
+		}
+
+		console.log(`Fetched ${models.length} tool-capable models from Requesty`);
+		return models;
+	} catch (error) {
+		console.error("Failed to fetch Requesty models:", error);
+		if (generatorOptions.strict) throw error;
+		return [];
+	}
+}
+
 function processZaiModels(data: ModelsDevCatalog): Model<Api>[] {
 	const variants = [
 		{
@@ -2549,9 +2630,10 @@ async function generateModels() {
 	const modelsDevModels = await loadModelsDevData();
 	const openRouterModels = await fetchOpenRouterModels();
 	const aiGatewayModels = await fetchAiGatewayModels();
+	const requestyModels = await fetchRequestyModels();
 
 	// Combine models (models.dev has priority)
-	const allModels = [...modelsDevModels, ...openRouterModels, ...aiGatewayModels].filter(
+	const allModels = [...modelsDevModels, ...openRouterModels, ...aiGatewayModels, ...requestyModels].filter(
 		(model) =>
 			!(model.provider === "xai" && XAI_BUILTIN_EXCLUDED_MODEL_IDS.has(model.id)) &&
 			!((model.provider === "opencode" || model.provider === "opencode-go") && model.id === "gpt-5.3-codex-spark"),

packages/ai/src/env-api-keys.ts

@@ -93,6 +93,7 @@ function getApiKeyEnvVars(provider: string): readonly string[] | undefined {
 		radius: "RADIUS_API_KEY",
 		openrouter: "OPENROUTER_API_KEY",
 		"vercel-ai-gateway": "AI_GATEWAY_API_KEY",
+		requesty: "REQUESTY_API_KEY",
 		zai: "ZAI_API_KEY",
 		"zai-coding-cn": "ZAI_CODING_CN_API_KEY",
 		mistral: "MISTRAL_API_KEY",

packages/ai/src/models.generated.ts

@@ -31,6 +31,7 @@ import { OPENROUTER_MODELS } from "./providers/openrouter.models.ts";
 import { QWEN_TOKEN_PLAN_MODELS } from "./providers/qwen-token-plan.models.ts";
 import { QWEN_TOKEN_PLAN_CN_MODELS } from "./providers/qwen-token-plan-cn.models.ts";
 import { QWEN_TOKEN_PLAN_INDIVIDUAL_MODELS } from "./providers/qwen-token-plan-individual.models.ts";
+import { REQUESTY_MODELS } from "./providers/requesty.models.ts";
 import { TOGETHER_MODELS } from "./providers/together.models.ts";
 import { VERCEL_AI_GATEWAY_MODELS } from "./providers/vercel-ai-gateway.models.ts";
 import { XAI_MODELS } from "./providers/xai.models.ts";
@@ -72,6 +73,7 @@ export const MODELS: {
 	readonly "qwen-token-plan": typeof QWEN_TOKEN_PLAN_MODELS;
 	readonly "qwen-token-plan-cn": typeof QWEN_TOKEN_PLAN_CN_MODELS;
 	readonly "qwen-token-plan-individual": typeof QWEN_TOKEN_PLAN_INDIVIDUAL_MODELS;
+	readonly "requesty": typeof REQUESTY_MODELS;
 	readonly "together": typeof TOGETHER_MODELS;
 	readonly "vercel-ai-gateway": typeof VERCEL_AI_GATEWAY_MODELS;
 	readonly "xai": typeof XAI_MODELS;
@@ -112,6 +114,7 @@ export const MODELS: {
 	"qwen-token-plan": QWEN_TOKEN_PLAN_MODELS,
 	"qwen-token-plan-cn": QWEN_TOKEN_PLAN_CN_MODELS,
 	"qwen-token-plan-individual": QWEN_TOKEN_PLAN_INDIVIDUAL_MODELS,
+	"requesty": REQUESTY_MODELS,
 	"together": TOGETHER_MODELS,
 	"vercel-ai-gateway": VERCEL_AI_GATEWAY_MODELS,
 	"xai": XAI_MODELS,

packages/ai/src/providers/all.ts

@@ -35,6 +35,7 @@ import { qwenTokenPlanProvider } from "./qwen-token-plan.ts";
 import { qwenTokenPlanCnProvider } from "./qwen-token-plan-cn.ts";
 import { qwenTokenPlanIndividualProvider } from "./qwen-token-plan-individual.ts";
 import { radiusProvider } from "./radius.ts";
+import { requestyProvider } from "./requesty.ts";
 import { togetherProvider } from "./together.ts";
 import { vercelAIGatewayProvider } from "./vercel-ai-gateway.ts";
 import { xaiProvider } from "./xai.ts";
@@ -119,6 +120,7 @@ export function builtinProviders(): Provider[] {
 		qwenTokenPlanCnProvider(),
 		qwenTokenPlanIndividualProvider(),
 		radiusProvider(),
+		requestyProvider(),
 		togetherProvider(),
 		vercelAIGatewayProvider(),
 		xaiProvider(),

packages/ai/src/providers/requesty.models.ts

@@ -0,0 +1,8 @@
+// This file is auto-generated by scripts/generate-models.ts
+// Do not edit manually - run 'npm run generate-models' to update
+
+import values from "./data/requesty.json" with { type: "json" };
+import { flattenModelCatalog, type ModelCatalog } from "../model-catalog.ts";
+
+export const REQUESTY_MODELS: ModelCatalog<typeof values, "requesty"> =
+	flattenModelCatalog("requesty", values);

packages/ai/src/providers/requesty.ts

@@ -0,0 +1,18 @@
+import { anthropicMessagesApi } from "../api/anthropic-messages.lazy.ts";
+import { openAICompletionsApi } from "../api/openai-completions.lazy.ts";
+import { envApiKeyAuth } from "../auth/helpers.ts";
+import { createProvider, type Provider } from "../models.ts";
+import { REQUESTY_MODELS } from "./requesty.models.ts";
+
+export function requestyProvider(): Provider<"anthropic-messages" | "openai-completions"> {
+	return createProvider({
+		id: "requesty",
+		name: "Requesty",
+		auth: { apiKey: envApiKeyAuth("Requesty API key", ["REQUESTY_API_KEY"]) },
+		models: Object.values(REQUESTY_MODELS),
+		api: {
+			"anthropic-messages": anthropicMessagesApi(),
+			"openai-completions": openAICompletionsApi(),
+		},
+	});
+}

packages/ai/src/types.ts

@@ -50,6 +50,7 @@ export type KnownProvider =
 	| "cerebras"
 	| "openrouter"
 	| "vercel-ai-gateway"
+	| "requesty"
 	| "zai"
 	| "zai-coding-cn"
 	| "mistral"

packages/ai/test/providers.test.ts

@@ -11,6 +11,7 @@ import { cloudflareAIGatewayProvider } from "../src/providers/cloudflare-ai-gate
 import { cloudflareWorkersAIProvider } from "../src/providers/cloudflare-workers-ai.ts";
 import { fauxAssistantMessage, fauxProvider } from "../src/providers/faux.ts";
 import { googleVertexProvider } from "../src/providers/google-vertex.ts";
+import { requestyProvider } from "../src/providers/requesty.ts";
 import type {
 	Api,
 	DeferredCancelOptions,
@@ -276,6 +277,29 @@ describe("builtin providers", () => {
 		).toMatchObject({ auth: {}, env: { AWS_PROFILE: "work" } });
 	});
 
+	it("routes Requesty models through the router's Anthropic and OpenAI-compatible endpoints", async () => {
+		const models = createModels({ authContext: fakeAuthContext({ REQUESTY_API_KEY: "rq-key" }) });
+		models.setProvider(requestyProvider());
+
+		expect(await models.getAuth("requesty")).toEqual({
+			auth: { apiKey: "rq-key" },
+			source: "REQUESTY_API_KEY",
+		});
+
+		const requestyModels = models.getModels("requesty");
+		expect(requestyModels.length).toBeGreaterThan(0);
+		for (const model of requestyModels) {
+			if (model.id.startsWith("anthropic/")) {
+				expect(model.api, model.id).toBe("anthropic-messages");
+				expect(model.baseUrl, model.id).toBe("https://router.requesty.ai");
+			} else {
+				expect(model.api, model.id).toBe("openai-completions");
+				expect(model.baseUrl, model.id).toBe("https://router.requesty.ai/v1");
+			}
+		}
+		expect(models.getModel("requesty", "anthropic/claude-sonnet-4-6")?.api).toBe("anthropic-messages");
+	});
+
 	it("reports bedrock as configured from ambient AWS credentials without an api key", async () => {
 		const models = createModels({ authContext: fakeAuthContext({ AWS_PROFILE: "dev" }) });
 		models.setProvider(amazonBedrockProvider());

packages/coding-agent/README.md

@@ -121,6 +121,7 @@ For each built-in provider, pi maintains a list of tool-capable models. Configur
 - xAI
 - OpenRouter
 - Vercel AI Gateway
+- Requesty
 - ZAI Coding Plan (Global)
 - ZAI Coding Plan (China)
 - OpenCode Zen

packages/coding-agent/docs/providers.md

@@ -84,6 +84,7 @@ pi
 | xAI | `XAI_API_KEY` | `xai` |
 | OpenRouter | `OPENROUTER_API_KEY` | `openrouter` |
 | Vercel AI Gateway | `AI_GATEWAY_API_KEY` | `vercel-ai-gateway` |
+| Requesty | `REQUESTY_API_KEY` | `requesty` |
 | ZAI Coding Plan (Global) | `ZAI_API_KEY` | `zai` |
 | ZAI Coding Plan (China) | `ZAI_CODING_CN_API_KEY` | `zai-coding-cn` |
 | OpenCode Zen | `OPENCODE_API_KEY` | `opencode` |

packages/coding-agent/src/cli/args.ts

@@ -406,6 +406,7 @@ ${chalk.bold("Environment Variables:")}
   BASETEN_API_KEY                  - Baseten API key
   OPENROUTER_API_KEY               - OpenRouter API key
   AI_GATEWAY_API_KEY               - Vercel AI Gateway API key
+  REQUESTY_API_KEY                 - Requesty API key
   ZAI_API_KEY                      - ZAI Coding Plan API key (Global)
   ZAI_CODING_CN_API_KEY            - ZAI Coding Plan API key (China)
   MISTRAL_API_KEY                  - Mistral API key

packages/coding-agent/src/core/model-resolver.ts

@@ -32,6 +32,7 @@ export const defaultModelPerProvider: Record<KnownProvider, string> = {
 	"github-copilot": "gpt-5.4",
 	openrouter: "moonshotai/kimi-k2.6",
 	"vercel-ai-gateway": "zai/glm-5.1",
+	requesty: "anthropic/claude-sonnet-4-6",
 	xai: "grok-4.6",
 	groq: "openai/gpt-oss-120b",
 	cerebras: "gpt-oss-120b",