diff --git a/python/google-adk/README.md b/python/google-adk/README.md
index 7d574e6..1fa5d6b 100644
--- a/python/google-adk/README.md
+++ b/python/google-adk/README.md
@@ -96,8 +96,8 @@ uv run pytest tests/ -v
 ## Switching to production mode
 
 1. Start an Agent Assembly gateway or use your SaaS workspace URL.
-2. Copy `.env.example` to `.env` and fill in your credentials (including a Gemini / Vertex AI key to drive a live model).
-3. Build a real `google.adk.agents.Agent` with your tools and run it via the ADK `Runner`.
+2. Copy `.env.example` to `.env` and fill in your credentials.
+3. Build a real `google.adk.agents.Agent` with your tools and run it via the ADK `Runner`, setting a Gemini / Vertex AI key to drive the live model.
 4. Run with gateway environment variables:
 
 ```bash
diff --git a/python/openai-agents-sdk/README.md b/python/openai-agents-sdk/README.md
index 40f9f22..a831a58 100644
--- a/python/openai-agents-sdk/README.md
+++ b/python/openai-agents-sdk/README.md
@@ -85,21 +85,19 @@ Running governed tool calls:
 uv run pytest tests/ -v
 ```
 
-## Switching to production mode (with real OpenAI API key)
+## Switching to production mode
 
 1. Start an Agent Assembly gateway or use your SaaS workspace URL.
 2. Copy `.env.example` to `.env` and fill in your credentials.
-3. Run with environment variables:
+3. Extend `main.py` to create a real `agents.Agent` with your `FunctionTool` instances, and set your `OPENAI_API_KEY`. Agent Assembly's `OpenAIAgentsPatch` intercepts every tool call automatically once `init_assembly()` has run.
+4. Run with gateway environment variables:
 
 ```bash
-OPENAI_API_KEY=sk-... \
 AGENT_ASSEMBLY_GATEWAY_URL=http://localhost:8080 \
 AGENT_ASSEMBLY_API_KEY=your-key \
 uv run python src/main.py
 ```
 
-When an `OPENAI_API_KEY` is set, you can extend `main.py` to create a real `agents.Agent` with your `FunctionTool` instances. Agent Assembly's `OpenAIAgentsPatch` intercepts every tool call automatically once `init_assembly()` has run.
-
 ## Troubleshooting
 
 | Problem | Fix |
diff --git a/python/openai-agents-sdk/src/main.py b/python/openai-agents-sdk/src/main.py
index cc5c7bc..049f807 100644
--- a/python/openai-agents-sdk/src/main.py
+++ b/python/openai-agents-sdk/src/main.py
@@ -7,9 +7,8 @@
 Run (offline mode, no API key required):
     uv run python src/main.py
 
-For production use with a real OpenAI API key and gateway:
-    OPENAI_API_KEY=sk-... AGENT_ASSEMBLY_GATEWAY_URL=http://localhost:8080 \\
-    uv run python src/main.py
+For production use, start the Agent Assembly gateway and update the gateway URL:
+    AGENT_ASSEMBLY_GATEWAY_URL=http://localhost:8080 uv run python src/main.py
 """
 from __future__ import annotations
 
diff --git a/python/smolagents-tool-policy/README.md b/python/smolagents-tool-policy/README.md
index d19acef..6e11a2f 100644
--- a/python/smolagents-tool-policy/README.md
+++ b/python/smolagents-tool-policy/README.md
@@ -102,8 +102,8 @@ The smoke suite includes a **negative control** (`test_negative_control_ungovern
 ## Switching to production mode
 
 1. Start an Agent Assembly gateway or use your SaaS workspace URL.
-2. Copy `.env.example` to `.env` and fill in your credentials (and a model provider key to drive a live smolagents agent loop).
-3. Build a real `smolagents.ToolCallingAgent` (or `CodeAgent`) with your tools and run it.
+2. Copy `.env.example` to `.env` and fill in your credentials.
+3. Build a real `smolagents.ToolCallingAgent` (or `CodeAgent`) with your tools and run it, setting a model provider key to drive the live agent loop.
 4. Run with gateway environment variables:
 
 ```bash