Open-Curiosity · Thibaultjaigu · Jun 24, 2026
diff --git a/README.md b/README.md
@@ -13,7 +13,7 @@ Gini is not just a chat box, CLI, messaging bot, or pile of tools. Chat is an in
 - [Memory](docs/memory.md): retain, recall, embeddings, reranking, review, and storage
 - [Skill Learning From Skills](docs/skill-learning.md): how Gini improves its own skills from task outcomes (two-tier reward, attribution, the daily review, the human gate)
 - [Runtime Capabilities](docs/runtime-capabilities.md): current CLI/API capability map and verification commands
-- [Model Providers](docs/providers/README.md): per-provider setup guides (credentials, prerequisites, CLI/web config) for OpenAI, Anthropic, Bedrock, Azure, OpenRouter, DeepSeek, Codex, and Local
+- [Model Providers](docs/providers/README.md): per-provider setup guides (credentials, prerequisites, CLI/web config) for OpenAI, Anthropic, Bedrock, Azure, OpenRouter, Requesty, DeepSeek, Codex, and Local
 - [Operations](docs/operations.md): install, start, stop, smoke, diagnostics, and cleanup
 - [Remote Access](docs/remote-access.md): tunnel modes and confirmation, plus a self-contained guide per tunnel provider — [Gini Relay](docs/remote-access/gini-relay.md), [Tailscale](docs/remote-access/tailscale.md), [ngrok](docs/remote-access/ngrok.md), [Cloudflare](docs/remote-access/cloudflare.md) — the same pages the app opens inline
 - [Releases](docs/releases.md): versioning, CHANGELOG conventions, and the release process
@@ -45,7 +45,7 @@ Gini's **runtime is the gateway**: a single Bun process per instance owns state
 - Authenticated localhost gateway and a Next.js + Tailwind + shadcn/ui control plane
 - Persistent chat, runs, tasks, approvals, traces, audit events, jobs, memories, and skills
 - Approval-gated file, terminal, and code tools
-- Provider support: Codex OAuth; OpenAI / Azure OpenAI / DeepSeek / OpenRouter API keys; the first-party Anthropic Claude API; Amazon Bedrock (model-agnostic Converse, AWS SigV4 — Claude, Nova, Llama, Mistral, DeepSeek); and any OpenAI-compatible local server
+- Provider support: Codex OAuth; OpenAI / Azure OpenAI / DeepSeek / OpenRouter / Requesty API keys; the first-party Anthropic Claude API; Amazon Bedrock (model-agnostic Converse, AWS SigV4 — Claude, Nova, Llama, Mistral, DeepSeek); and any OpenAI-compatible local server
 - Local embeddings, reranking, and voice-message speech-to-text by default
 - Parallel instances with isolated state, ports, and logs
 
@@ -57,7 +57,7 @@ See the [Whitepaper](docs/whitepaper.md) and [Architecture Overview](docs/archit
 curl -fsSL https://raw.githubusercontent.com/Lilac-Labs/gini-agent/main/scripts/install.sh | bash
 ```
 
-On macOS the installer enables autostart (per-user LaunchAgents for the runtime and webapp), waits for the webapp to come up, and opens the `/setup` page in your browser. The form offers the full provider catalog — OpenAI, Codex, Anthropic, Amazon Bedrock, Azure OpenAI, OpenRouter, DeepSeek, and Local — and prompts for whatever the one you pick needs (an API key, the AWS access key pair for Bedrock, the resource endpoint for Azure, or your existing `codex login` auth). Save it and you land on the running app. The runtime stays alive across reboots and crashes until you explicitly run `gini stop` or `gini autostart disable`.
+On macOS the installer enables autostart (per-user LaunchAgents for the runtime and webapp), waits for the webapp to come up, and opens the `/setup` page in your browser. The form offers the full provider catalog — OpenAI, Codex, Anthropic, Amazon Bedrock, Azure OpenAI, OpenRouter, Requesty, DeepSeek, and Local — and prompts for whatever the one you pick needs (an API key, the AWS access key pair for Bedrock, the resource endpoint for Azure, or your existing `codex login` auth). Save it and you land on the running app. The runtime stays alive across reboots and crashes until you explicitly run `gini stop` or `gini autostart disable`.
 
 If the browser doesn't open automatically (or you want to navigate manually), run `gini status` to print the actual web URL. The installed `default` instance always lives at `:7777`; other instances get hash-derived ports, so check `gini status` rather than guessing. The installer also prints the URL right before opening the browser.
 
@@ -114,6 +114,7 @@ Run `gini setup` for an interactive picker, or configure directly:
 gini provider set codex gpt-5.5            # Codex OAuth (reads ~/.codex/auth.json)
 gini provider set openai gpt-5.4-mini      # uses $OPENAI_API_KEY
 gini provider set openrouter <model>       # uses $OPENROUTER_API_KEY
+gini provider set requesty <model>         # uses $REQUESTY_API_KEY
 gini provider set local <model> --base-url http://127.0.0.1:8000/v1
 gini provider set anthropic claude-opus-4-8 # first-party Claude API, uses $ANTHROPIC_API_KEY
 # Amazon Bedrock: model-agnostic Converse, SigV4-signed with AWS keys you enter via the web form or `gini setup`
@@ -124,11 +125,11 @@ gini provider set azure gpt-4o \
   --deployment <deployment> --api-version 2024-10-21 --auth-scheme api-key  # uses $AZURE_OPENAI_API_KEY
 ```
 
-For step-by-step setup of each provider — getting credentials, installing any prerequisite tooling (Ollama, …), and configuring it in the CLI or web — see the per-provider guides: [OpenAI](docs/providers/openai.md), [Anthropic](docs/providers/anthropic.md), [Amazon Bedrock](docs/providers/bedrock.md), [Azure OpenAI](docs/providers/azure.md), [OpenRouter](docs/providers/openrouter.md), [DeepSeek](docs/providers/deepseek.md), [Codex](docs/providers/codex.md), and [Local](docs/providers/local.md). The [providers index](docs/providers/README.md) lists them all with their auth model at a glance.
+For step-by-step setup of each provider — getting credentials, installing any prerequisite tooling (Ollama, …), and configuring it in the CLI or web — see the per-provider guides: [OpenAI](docs/providers/openai.md), [Anthropic](docs/providers/anthropic.md), [Amazon Bedrock](docs/providers/bedrock.md), [Azure OpenAI](docs/providers/azure.md), [OpenRouter](docs/providers/openrouter.md), [Requesty](docs/providers/requesty.md), [DeepSeek](docs/providers/deepseek.md), [Codex](docs/providers/codex.md), and [Local](docs/providers/local.md). The [providers index](docs/providers/README.md) lists them all with their auth model at a glance.
 
 The `local` provider works with any OpenAI-compatible server (oMLX, vLLM, LM Studio, llama.cpp). The `azure` provider targets an Azure OpenAI resource: set `--base-url` to `https://<resource>.openai.azure.com` and pick a deployment; `--api-version` defaults to a GA value and `--auth-scheme` defaults to `api-key` (a resource key), with `bearer` available for an Entra token. API keys are read from environment variables, and Codex OAuth is read from `~/.codex/auth.json` (or `CODEX_AUTH_JSON`) — nothing is written to Gini config. Run `gini --help` for the full flag set, or see [provider-extra-body.md](docs/adr/provider-extra-body.md) for the `--extra-body` contract and [Azure OpenAI As A First-Class Provider](docs/adr/azure-provider.md) for the Azure routing contract. When a credential fails mid-chat, see [Codex re-authentication](docs/providers/codex.md#re-authentication) and [Provider Re-Authentication Guidance](docs/adr/provider-reauth-guidance.md).
 
-`gini setup`'s interactive picker covers every provider — OpenAI, Codex, Anthropic, Amazon Bedrock, Azure OpenAI, OpenRouter, DeepSeek, and Local — prompting for whatever each one needs (an API key, the AWS access key + secret for Bedrock, the resource endpoint and deployment for Azure, the base URL for Local). `gini provider set …` (above) and the web **Settings → Add provider** form remain available for scripted or non-interactive configuration.
+`gini setup`'s interactive picker covers every provider — OpenAI, Codex, Anthropic, Amazon Bedrock, Azure OpenAI, OpenRouter, Requesty, DeepSeek, and Local — prompting for whatever each one needs (an API key, the AWS access key + secret for Bedrock, the resource endpoint and deployment for Azure, the base URL for Local). `gini provider set …` (above) and the web **Settings → Add provider** form remain available for scripted or non-interactive configuration.
 
 ## Parallel Instances
 

diff --git a/docs/providers/README.md b/docs/providers/README.md
@@ -11,6 +11,7 @@ provider in Gini (both the CLI and the web Add Provider form).
 | Amazon Bedrock | AWS SigV4 | AWS access key + secret you enter (stored in `~/.gini/secrets.env`) | [bedrock.md](bedrock.md) |
 | Azure OpenAI | API key / Entra | `AZURE_OPENAI_API_KEY` | [azure.md](azure.md) |
 | OpenRouter | API key | `OPENROUTER_API_KEY` | [openrouter.md](openrouter.md) |
+| Requesty | API key | `REQUESTY_API_KEY` | [requesty.md](requesty.md) |
 | DeepSeek | API key | `DEEPSEEK_API_KEY` | [deepseek.md](deepseek.md) |
 | Codex (OpenAI OAuth) | OAuth / CLI | `~/.codex/auth.json` (no key) | [codex.md](codex.md) |
 | Local (OpenAI-compatible) | none / optional key | `GINI_LOCAL_API_KEY` (optional) | [local.md](local.md) |

diff --git a/docs/providers/requesty.md b/docs/providers/requesty.md
@@ -0,0 +1,50 @@
+# Requesty
+
+Requesty is an OpenAI-compatible API-key provider that routes a single key to
+models from many vendors. Gini talks to `https://router.requesty.ai/v1` and
+authenticates with a Bearer key.
+
+## Step 1 — Get an API key
+
+1. Sign in at [requesty.ai](https://requesty.ai/).
+2. Create a key on the [API keys page](https://app.requesty.ai/api-keys) and copy it.
+3. Add credit (Requesty is pay-as-you-go and bills per model). See the
+   [Requesty docs](https://docs.requesty.ai/) for the model catalog.
+
+See the [Requesty docs](https://docs.requesty.ai/) for the full API reference.
+
+## Step 2 — Set the key
+
+Gini reads the key from the `REQUESTY_API_KEY` environment variable. Set it in
+your shell or in `~/.gini/secrets.env` for persistence:
+
+```bash
+# ~/.gini/secrets.env  (created mode 0600)
+REQUESTY_API_KEY=rqsty-sk-...
+```
+
+The web Add Provider form writes this for you.
+
+## Step 3 — Configure the provider in Gini
+
+### CLI
+
+```bash
+gini provider set requesty openai/gpt-4o-mini
+```
+
+Requesty uses `provider/model` slugs (e.g. `openai/gpt-4o-mini`,
+`anthropic/claude-opus-4-8`). The base URL defaults to
+`https://router.requesty.ai/v1`; override it only for a proxy with `--base-url`.
+
+### Web
+
+Open **Settings → Add provider → Requesty**, paste the key, and pick or type a
+model slug.
+
+## Re-authentication
+
+Requesty is an API-key provider, so a credential failure surfaces Requesty's own
+message and links to **Settings → Providers** to paste a new key. Rotate keys on
+the [API keys page](https://app.requesty.ai/api-keys). See ADR
+[provider-reauth-guidance.md](../adr/provider-reauth-guidance.md).
diff --git a/src/cli/commands/provider.ts b/src/cli/commands/provider.ts
@@ -44,6 +44,7 @@ export async function provider(ctx: CliContext): Promise<void> {
       name !== "openai" &&
       name !== "codex" &&
       name !== "openrouter" &&
+      name !== "requesty" &&
       name !== "local" &&
       name !== "deepseek" &&
       name !== "anthropic" &&
@@ -169,7 +170,7 @@ export async function provider(ctx: CliContext): Promise<void> {
 
     config.provider = normalizeProvider({
       name,
-      model: model ?? (name === "echo" ? "gini-echo-v0" : name === "codex" ? "gpt-5.5" : name === "openrouter" ? "openrouter/auto" : name === "local" ? "local/default" : name === "deepseek" ? "deepseek-v4-flash" : name === "anthropic" ? "claude-opus-4-8" : name === "bedrock" ? "us.anthropic.claude-opus-4-8" : name === "azure" ? "gpt-5.5" : "gpt-5.4-mini"),
+      model: model ?? (name === "echo" ? "gini-echo-v0" : name === "codex" ? "gpt-5.5" : name === "openrouter" ? "openrouter/auto" : name === "requesty" ? "openai/gpt-4o-mini" : name === "local" ? "local/default" : name === "deepseek" ? "deepseek-v4-flash" : name === "anthropic" ? "claude-opus-4-8" : name === "bedrock" ? "us.anthropic.claude-opus-4-8" : name === "azure" ? "gpt-5.5" : "gpt-5.4-mini"),
       ...(baseUrl ? { baseUrl } : {}),
       ...(apiKeyEnv ? { apiKeyEnv } : {}),
       ...(extraBody ? { extraBody } : {}),

diff --git a/src/cli/commands/setup.ts b/src/cli/commands/setup.ts
@@ -123,7 +123,7 @@ export interface ProviderExtraConfig {
 }
 
 export interface ProviderModule {
-  id: "openai" | "codex" | "anthropic" | "bedrock" | "azure" | "openrouter" | "deepseek" | "local";
+  id: "openai" | "codex" | "anthropic" | "bedrock" | "azure" | "openrouter" | "requesty" | "deepseek" | "local";
   kind: ProviderKind;
   label: string;
   description: string;
@@ -231,6 +231,16 @@ const openrouterProvider: ProviderModule = apiKeyProvider({
   keyHint: "sk-or-"
 });
 
+const requestyProvider: ProviderModule = apiKeyProvider({
+  id: "requesty",
+  label: "Requesty",
+  description: "Multi-model router — rqsty-sk-...",
+  apiKeyEnv: "REQUESTY_API_KEY",
+  defaultModel: "openai/gpt-4o-mini",
+  suggestedModels: ["openai/gpt-4o-mini", "openai/gpt-4o", "anthropic/claude-opus-4-8"],
+  keyHint: "rqsty-sk-"
+});
+
 const deepseekProvider: ProviderModule = apiKeyProvider({
   id: "deepseek",
   label: "DeepSeek",
@@ -460,6 +470,7 @@ const PROVIDERS: ProviderModule[] = [
   bedrockProvider,
   azureProvider,
   openrouterProvider,
+  requestyProvider,
   deepseekProvider,
   localProvider
 ];
@@ -528,6 +539,7 @@ const AUTO_CONFIGURABLE: ProviderModule[] = [
   openaiProvider,
   anthropicProvider,
   openrouterProvider,
+  requestyProvider,
   deepseekProvider,
   bedrockProvider
 ];
@@ -722,6 +734,7 @@ export const __testing = {
   bedrockProvider,
   azureProvider,
   openrouterProvider,
+  requestyProvider,
   deepseekProvider,
   localProvider,
   PROVIDERS,

diff --git a/src/cli/output.ts b/src/cli/output.ts
@@ -112,16 +112,16 @@ Usage:
   bun run gini notifications list|queue|send|ack
   bun run gini promotions list|propose|approve|reject
   bun run gini snapshots list|create|restore
-  bun run gini provider show|catalog|set echo|openai|codex|openrouter|local|deepseek|azure [model]
+  bun run gini provider show|catalog|set echo|openai|codex|openrouter|requesty|local|deepseek|azure [model]
                   [--base-url <url>] [--api-key-env <NAME>] [--extra-body <JSON>]
                   [--api-version <V>] [--deployment <NAME>] [--auth-scheme bearer|api-key]
                   --base-url and --api-key-env work for every OpenAI-compatible
-                  provider (local / openai / openrouter / deepseek / azure —
+                  provider (local / openai / openrouter / requesty / deepseek / azure —
                   point at servers like oMLX, vLLM, LM Studio) AND for codex
                   (override the backend URL or auth-file env var). --extra-body
                   forwards server-specific request fields like
                   \`chat_template_kwargs\` and applies to the chat-completions
-                  calls of local / openai / openrouter / deepseek / azure;
+                  calls of local / openai / openrouter / requesty / deepseek / azure;
                   codex (/responses) and echo ignore it. --api-version,
                   --deployment, and --auth-scheme configure the azure provider
                   (Azure OpenAI): set --base-url to https://<resource>.openai.azure.com;

diff --git a/src/embeddings.ts b/src/embeddings.ts
@@ -378,7 +378,7 @@ async function embedOpenAIBatch(config: RuntimeConfig, texts: string[]): Promise
 // the request and leak the chat key to that host (e.g. the Anthropic key to
 // api.anthropic.com/embeddings). Those fall back to the canonical OpenAI
 // endpoint + OPENAI_API_KEY instead.
-const OPENAI_EMBEDDING_COMPATIBLE = new Set(["openai", "openrouter", "deepseek", "local"]);
+const OPENAI_EMBEDDING_COMPATIBLE = new Set(["openai", "openrouter", "requesty", "deepseek", "local"]);
 
 function openAIEmbeddingBaseUrl(config: RuntimeConfig): string {
   const reuse = OPENAI_EMBEDDING_COMPATIBLE.has(config.provider.name) && config.provider.baseUrl;

diff --git a/src/model-routes.ts b/src/model-routes.ts
@@ -40,7 +40,7 @@ export const MODEL_ALIASES: Record<string, Record<string, { id: string; qualifie
 // revisions and features first), codex's OAuth bundle beats the metered
 // clouds, and openrouter/local are deliberate opt-ins that should never
 // win a tie. Unknown providers sort last in catalog order.
-const ROUTE_PRIORITY = ["openai", "anthropic", "deepseek", "codex", "azure", "bedrock", "openrouter", "local"] as const;
+const ROUTE_PRIORITY = ["openai", "anthropic", "deepseek", "codex", "azure", "bedrock", "openrouter", "requesty", "local"] as const;
 
 function routePriority(provider: ProviderCatalogItem["name"]): number {
   const index = (ROUTE_PRIORITY as readonly string[]).indexOf(provider);

diff --git a/src/provider-capabilities.ts b/src/provider-capabilities.ts
@@ -135,6 +135,10 @@ export function resolveProviderContextWindowTokens(provider: ProviderConfig): nu
       return Math.min(CODEX_BACKEND_CONTEXT_WINDOW_TOKENS, openaiContextWindowTokens(model));
     case "openrouter":
       return openrouterContextWindowTokens(model);
+    case "requesty":
+      // Requesty uses the same <vendor>/<model> slug scheme as OpenRouter,
+      // so the per-vendor context-window lookup applies unchanged.
+      return openrouterContextWindowTokens(model);
     case "deepseek":
       return deepseekContextWindowTokens(model);
     case "anthropic":
@@ -187,6 +191,9 @@ export function resolveProviderModality(provider: ProviderConfig): ProviderModal
         : { vision: false, nativeDocs: false };
     case "openrouter":
       return openrouterModality(model);
+    case "requesty":
+      // Same <vendor>/<model> slug scheme as OpenRouter; reuse its modality map.
+      return openrouterModality(model);
     case "deepseek":
       // Confirmed text-only API — no image/file content part.
       return { vision: false, nativeDocs: false };

diff --git a/src/provider.ts b/src/provider.ts
@@ -136,6 +136,7 @@ export function providerHealth(config: RuntimeConfig) {
 const PROVIDER_API_KEY_ENV: Record<string, string> = {
   openai: "OPENAI_API_KEY",
   openrouter: "OPENROUTER_API_KEY",
+  requesty: "REQUESTY_API_KEY",
   deepseek: "DEEPSEEK_API_KEY",
   local: "GINI_LOCAL_API_KEY",
   anthropic: "ANTHROPIC_API_KEY",
@@ -376,6 +377,16 @@ export function providerCatalog(): ProviderCatalogItem[] {
       capabilities: ["chat-completions", "model-routing"],
       costHint: "external"
     },
+    {
+      id: "requesty",
+      name: "requesty",
+      displayName: "Requesty Compatible",
+      baseUrl: "https://router.requesty.ai/v1",
+      auth: "env",
+      models: ["openai/gpt-4o-mini"],
+      capabilities: ["chat-completions", "model-routing"],
+      costHint: "external"
+    },
     {
       id: "deepseek",
       name: "deepseek",
@@ -429,6 +440,8 @@ export function providerDisplayLabel(name: ProviderName): string {
       return "OpenAI";
     case "openrouter":
       return "OpenRouter";
+    case "requesty":
+      return "Requesty";
     case "deepseek":
       return "DeepSeek";
     case "anthropic":
@@ -3082,6 +3095,7 @@ export async function generateTaskSummary(
     }
     if (
       provider.name === "openrouter" ||
+      provider.name === "requesty" ||
       provider.name === "local" ||
       provider.name === "deepseek" ||
       // Azure OpenAI exposes deployment-scoped chat/completions, not the flat
@@ -3225,6 +3239,7 @@ export async function generateStructured<T>(
   // many compat providers reject the field. Validator re-checks shape.
   if (
     provider.name === "openrouter" ||
+    provider.name === "requesty" ||
     provider.name === "local" ||
     provider.name === "openai" ||
     provider.name === "deepseek" ||
@@ -3436,6 +3451,15 @@ export function normalizeProvider(provider: ProviderConfig): ProviderConfig {
       ...(provider.extraBody ? { extraBody: provider.extraBody } : {})
     };
   }
+  if (provider.name === "requesty") {
+    return {
+      name: "requesty",
+      model: provider.model || "openai/gpt-4o-mini",
+      baseUrl: pickBaseUrl(provider.baseUrl, "https://router.requesty.ai/v1"),
+      apiKeyEnv: provider.apiKeyEnv ?? "REQUESTY_API_KEY",
+      ...(provider.extraBody ? { extraBody: provider.extraBody } : {})
+    };
+  }
   if (provider.name === "local") {
     return {
       name: "local",
@@ -4329,6 +4353,7 @@ function resolveBaseUrl(baseUrl: string | undefined, fallback: string): string {
 function defaultBaseUrl(provider: ProviderConfig): string {
   if (provider.name === "codex") return DEFAULT_CODEX_BASE_URL;
   if (provider.name === "openrouter") return "https://openrouter.ai/api/v1";
+  if (provider.name === "requesty") return "https://router.requesty.ai/v1";
   if (provider.name === "local") return "http://127.0.0.1:11434/v1";
   if (provider.name === "deepseek") return DEFAULT_DEEPSEEK_BASE_URL;
   if (provider.name === "anthropic") return DEFAULT_ANTHROPIC_BASE_URL;