diff --git a/docs/src/app/api/page.tsx b/docs/src/app/api/page.tsx
index 58f6346e..ddca2671 100644
--- a/docs/src/app/api/page.tsx
+++ b/docs/src/app/api/page.tsx
@@ -54,15 +54,16 @@ print(result.response)`} />
               <span className="text-xs px-2 py-1 rounded-md bg-blue-100 text-blue-700 font-mono">default: "openai"</span>
             </div>
             <p className="text-muted-foreground mb-4">LM provider to use for completions.</p>
-            <Table 
+            <Table
               headers={["Value", "Provider"]}
               rows={[
                 [<code key="1" className="text-sm">"openai"</code>, "OpenAI API"],
                 [<code key="2" className="text-sm">"anthropic"</code>, "Anthropic API"],
-                [<code key="3" className="text-sm">"portkey"</code>, "Portkey AI gateway"],
-                [<code key="4" className="text-sm">"openrouter"</code>, "OpenRouter"],
-                [<code key="5" className="text-sm">"litellm"</code>, "LiteLLM (multi-provider)"],
-                [<code key="6" className="text-sm">"vllm"</code>, "Local vLLM server"],
+                [<code key="3" className="text-sm">"vercel"</code>, "Vercel AI Gateway"],
+                [<code key="4" className="text-sm">"portkey"</code>, "Portkey AI gateway"],
+                [<code key="5" className="text-sm">"openrouter"</code>, "OpenRouter"],
+                [<code key="6" className="text-sm">"litellm"</code>, "LiteLLM (multi-provider)"],
+                [<code key="7" className="text-sm">"vllm"</code>, "Local vLLM server"],
               ]}
             />
           </div>
@@ -81,6 +82,12 @@ backend_kwargs={
     "model_name": "gpt-5-mini",
 }
 
+# Vercel AI Gateway
+backend_kwargs={
+    "api_key": "...",
+    "model_name": "openai/gpt-5.2",  # Format: creator/model-name
+}
+
 # vLLM (local)
 backend_kwargs={
     "base_url": "http://localhost:8000/v1",
diff --git a/docs/src/app/backends/page.tsx b/docs/src/app/backends/page.tsx
index 6a0708aa..1808bddd 100644
--- a/docs/src/app/backends/page.tsx
+++ b/docs/src/app/backends/page.tsx
@@ -7,7 +7,7 @@ export default function BackendsPage() {
       
       <p className="text-muted-foreground mb-6">
         <p>
-          RLMs natively support a wide range of language model providers, including <code>OpenAI</code>, <code>Anthropic</code>, <code>Portkey</code>, <code>OpenRouter</code>, and <code>LiteLLM</code>. Additional providers can be supported with minimal effort. The <code>backend_kwargs</code> are named arguments passed directly to the backend client.
+          RLMs natively support a wide range of language model providers, including <code>OpenAI</code>, <code>Anthropic</code>, <code>Vercel AI Gateway</code>, <code>Portkey</code>, <code>OpenRouter</code>, and <code>LiteLLM</code>. Additional providers can be supported with minimal effort. The <code>backend_kwargs</code> are named arguments passed directly to the backend client.
         </p>
       </p>
 
@@ -64,6 +64,23 @@ export default function BackendsPage() {
 
       <hr className="my-8 border-border" />
 
+      <h2 className="text-2xl font-semibold mb-4">Vercel AI Gateway</h2>
+      <p className="text-muted-foreground mb-4">
+        <a href="https://vercel.com/docs/infrastructure/ai-gateway" className="text-primary underline font-medium" target="_blank" rel="noopener noreferrer">Vercel AI Gateway</a> provides a unified endpoint for accessing multiple AI providers with built-in caching and analytics.
+      </p>
+      <CodeBlock code={`rlm = RLM(
+    backend="vercel",
+    backend_kwargs={
+        "api_key": os.getenv("AI_GATEWAY_API_KEY"),
+        "model_name": "openai/gpt-5.2",  # Format: creator/model-name
+    },
+)`} />
+      <p className="text-muted-foreground mt-4">
+        Vercel AI Gateway provides OpenAI-compatible endpoints for multiple AI providers. Use the <code className="px-1.5 py-0.5 rounded bg-muted text-foreground text-sm">creator/model-name</code> format (e.g., <code className="px-1.5 py-0.5 rounded bg-muted text-foreground text-sm">openai/gpt-5.2</code>, <code className="px-1.5 py-0.5 rounded bg-muted text-foreground text-sm">anthropic/claude-sonnet-4.5</code>). Cost tracking works correctly through the standard usage API.
+      </p>
+
+      <hr className="my-8 border-border" />
+
       <h2 className="text-2xl font-semibold mb-4">LiteLLM</h2>
       <p className="text-muted-foreground mb-4">
         <a href="https://docs.litellm.ai/docs/" className="text-primary underline font-medium" target="_blank" rel="noopener noreferrer">LiteLLM</a> is a universal interface for 100+ model providers, with support for local models and custom endpoints.
diff --git a/docs/src/app/page.tsx b/docs/src/app/page.tsx
index 66d3baf4..b1a8a41a 100644
--- a/docs/src/app/page.tsx
+++ b/docs/src/app/page.tsx
@@ -96,6 +96,7 @@ uv pip install -e . --extra modal`} />
         <TabsList>
           <TabsTrigger value="openai">OpenAI</TabsTrigger>
           <TabsTrigger value="anthropic">Anthropic</TabsTrigger>
+          <TabsTrigger value="vercel">Vercel AI Gateway</TabsTrigger>
           <TabsTrigger value="portkey">Portkey</TabsTrigger>
         </TabsList>
         <TabsContent value="openai">
@@ -127,6 +128,22 @@ rlm = RLM(
     verbose=False,  # print to logs
 )
 
+result = rlm.completion("Calculate 2^(2^(2^2)) using Python.")
+print(result.response)`} />
+        </TabsContent>
+        <TabsContent value="vercel">
+          <CodeBlock code={`import os
+from rlm import RLM
+
+rlm = RLM(
+    backend="vercel",
+    backend_kwargs={
+        "api_key": os.getenv("AI_GATEWAY_API_KEY"),
+        "model_name": "openai/gpt-5.2",  # Format: creator/model-name
+    },
+    verbose=False,  # print to logs
+)
+
 result = rlm.completion("Calculate 2^(2^(2^2)) using Python.")
 print(result.response)`} />
         </TabsContent>
diff --git a/rlm/clients/__init__.py b/rlm/clients/__init__.py
index 71606268..0eb8a226 100644
--- a/rlm/clients/__init__.py
+++ b/rlm/clients/__init__.py
@@ -36,6 +36,11 @@ def get_client(
 
         backend_kwargs.setdefault("base_url", "https://openrouter.ai/api/v1")
         return OpenAIClient(**backend_kwargs)
+    elif backend == "vercel":
+        from rlm.clients.openai import OpenAIClient
+
+        backend_kwargs.setdefault("base_url", "https://ai-gateway.vercel.sh/v1")
+        return OpenAIClient(**backend_kwargs)
     elif backend == "litellm":
         from rlm.clients.litellm import LiteLLMClient
 
@@ -46,5 +51,5 @@ def get_client(
         return AnthropicClient(**backend_kwargs)
     else:
         raise ValueError(
-            f"Unknown backend: {backend}. Supported backends: ['openai', 'vllm', 'portkey', 'openrouter', 'litellm', 'anthropic']"
+            f"Unknown backend: {backend}. Supported backends: ['openai', 'vllm', 'portkey', 'openrouter', 'vercel', 'litellm', 'anthropic']"
         )
diff --git a/rlm/core/types.py b/rlm/core/types.py
index ad167628..0807139b 100644
--- a/rlm/core/types.py
+++ b/rlm/core/types.py
@@ -2,7 +2,7 @@
 from types import ModuleType
 from typing import Any, Literal
 
-ClientBackend = Literal["openai", "portkey", "openrouter", "vllm", "litellm", "anthropic"]
+ClientBackend = Literal["openai", "portkey", "openrouter", "vllm", "litellm", "anthropic", "vercel"]
 EnvironmentType = Literal["local", "prime", "modal"]