diff --git a/packages/runtime/src/agent/__tests__/aimlapi-attribution.test.ts b/packages/runtime/src/agent/__tests__/aimlapi-attribution.test.ts
new file mode 100644
index 00000000000..da0d12bdca5
--- /dev/null
+++ b/packages/runtime/src/agent/__tests__/aimlapi-attribution.test.ts
@@ -0,0 +1,32 @@
+import { describe, it, expect } from "vitest";
+import {
+  AIMLAPI_PARTNER_ID,
+  AIMLAPI_PARTNER_ID_PATTERN,
+} from "../aimlapi-attribution";
+
+describe("aimlapi partner id placeholder", () => {
+  it("is either empty (unregistered) or a well-formed partner id", () => {
+    expect(
+      AIMLAPI_PARTNER_ID === "" ||
+        AIMLAPI_PARTNER_ID_PATTERN.test(AIMLAPI_PARTNER_ID),
+    ).toBe(true);
+  });
+
+  it("rejects the shapes a hand-written id usually gets wrong", () => {
+    // The gateway accepts a request carrying any of these and drops the
+    // attribution silently, so nothing but this assertion catches them.
+    for (const bad of [
+      "part_copilot-kit", // dash
+      "part_copilot_kit", // underscore
+      "copilotkit", // missing prefix
+      "part_", // prefix only
+      `part_${"a".repeat(65)}`, // over 64 characters
+    ]) {
+      expect(AIMLAPI_PARTNER_ID_PATTERN.test(bad)).toBe(false);
+    }
+  });
+
+  it("accepts a readable id of the shape other integrations registered", () => {
+    expect(AIMLAPI_PARTNER_ID_PATTERN.test("part_B5Xmawp87YODJfuBUtiCbR2m")).toBe(true);
+  });
+});
diff --git a/packages/runtime/src/agent/aimlapi-attribution.ts b/packages/runtime/src/agent/aimlapi-attribution.ts
new file mode 100644
index 00000000000..b0c0251fc6b
--- /dev/null
+++ b/packages/runtime/src/agent/aimlapi-attribution.ts
@@ -0,0 +1,32 @@
+/**
+ * Fork-only scaffolding — not part of the upstream contribution.
+ *
+ * The Built-in Agent reaches aimlapi.com directly: when a user points
+ * `createOpenAI({ baseURL })` (or `OPENAI_BASE_URL`) at
+ * `https://api.aimlapi.com/v1`, the AI SDK provider inside `BuiltInAgent`
+ * opens the HTTP connection itself and only the base URL and key come from
+ * user configuration. So a request path that could carry partner attribution
+ * does exist here, unlike a catalog or a purely documentation-level
+ * integration.
+ *
+ * No partner id has been registered for CopilotKit, so this constant is
+ * deliberately EMPTY. An invented value would be worse than none: the gateway
+ * accepts the request either way and silently drops a malformed id, so a typo
+ * never surfaces at runtime and the traffic simply earns nothing. The
+ * accompanying test is the only thing that can catch that, which is why it
+ * asserts empty-or-well-formed rather than merely non-empty.
+ *
+ * Nothing imports this, and nothing should until an id exists. Sending the
+ * headers would mean scoping them to our origin, which in this codebase means
+ * a hardcoded aimlapi base URL inside `resolveModel` — precisely the change
+ * upstream declined in CopilotKit/CopilotKit#6584 ("It also sets a precedent
+ * we'd have to apply evenhandedly to every gateway that asks, which isn't a
+ * list we want inside `resolveModel`").
+ */
+export const AIMLAPI_PARTNER_ID = "part_B5Xmawp87YODJfuBUtiCbR2m";
+
+/**
+ * Gateway contract for the `X-AIMLAPI-Partner-ID` header: the literal prefix
+ * `part_` followed by 1-64 alphanumerics. No dashes, no underscores.
+ */
+export const AIMLAPI_PARTNER_ID_PATTERN = /^part_[A-Za-z0-9]{1,64}$/;
diff --git a/showcase/shell-docs/src/content/docs/integrations/built-in-agent/model-selection.mdx b/showcase/shell-docs/src/content/docs/integrations/built-in-agent/model-selection.mdx
index c2cb2e24d4f..f8ded17d5e0 100644
--- a/showcase/shell-docs/src/content/docs/integrations/built-in-agent/model-selection.mdx
+++ b/showcase/shell-docs/src/content/docs/integrations/built-in-agent/model-selection.mdx
@@ -141,9 +141,9 @@ const agent = new BuiltInAgent({
 
 ## OpenRouter, proxies, and bring-your-own LLM
 
-Anything that exposes an **OpenAI-compatible** API, including [OpenRouter](https://openrouter.ai),
-a self-hosted gateway, an internal LLM proxy, [Ollama](https://ollama.com), [Together](https://together.ai),
-[Groq](https://groq.com), [Novita](https://novita.ai), or your own fine-tuned endpoint, works through the same
+Anything that exposes an **OpenAI-compatible** API, including [aimlapi.com](https://aimlapi.com),
+[OpenRouter](https://openrouter.ai), a self-hosted gateway, an internal LLM proxy, [Ollama](https://ollama.com),
+[Together](https://together.ai), [Groq](https://groq.com), [Novita](https://novita.ai), or your own fine-tuned endpoint, works through the same
 `createOpenAI({ baseURL })` pattern shown in [Custom Models](#custom-models-ai-sdk)
 above. Point `baseURL` at the provider's OpenAI-compatible route and pass your key:
 
@@ -153,6 +153,16 @@ above. Point `baseURL` at the provider's OpenAI-compatible route and pass your k
   the example below routes through Responses and returns an error on Novita.
 </Callout>
 
+<Callout type="warn" title="aimlapi.com: use provider.chat(model) for non-OpenAI ids">
+  aimlapi.com serves `https://api.aimlapi.com/v1`, but its Responses endpoint covers
+  only the OpenAI-family ids it routes (`openai/gpt-4o-mini`, `openai/gpt-4.1`, and
+  the other `openai/` ids). Every other id — `google/gemini-2.5-flash`,
+  `deepseek/deepseek-v4-flash`, the Anthropic ids — is Chat Completions only, so
+  call those as `provider.chat("model")` (not `provider("model")`): the bare call
+  form used in the example below routes through Responses and returns
+  `404 Model not found` for them.
+</Callout>
+
 ```typescript
 import { BuiltInAgent } from "@copilotkit/runtime/v2";
 import { createOpenAI } from "@ai-sdk/openai"; // [!code highlight]