openrouter-openai-compat
Migrate from OpenAI to OpenRouter with minimal code changes. Use when switching to OpenRouter or maintaining dual compatibility. Triggers: 'openrouter openai compatible', 'openrouter drop-in', 'openai to openrouter', 'openrouter migration'.
Allowed Tools
Provided by Plugin
openrouter-pack
Flagship+ skill pack for OpenRouter - 30 skills for multi-model routing, fallbacks, and LLM gateway mastery
Installation
This skill is included in the openrouter-pack plugin:
/plugin install openrouter-pack@claude-code-plugins-plus
Click to copy
Instructions
OpenRouter OpenAI Compatibility
Overview
OpenRouter implements the OpenAI Chat Completions API specification (/v1/chat/completions). Existing OpenAI SDK code works with OpenRouter by changing two values: baseurl and apikey. This gives you access to 400+ models from all providers through the same SDK interface.
The Two-Line Migration
Python (Before)
from openai import OpenAI
client = OpenAI(api_key=os.environ["OPENAI_API_KEY"]) # OpenAI direct
response = client.chat.completions.create(
model="gpt-4o",
messages=[{"role": "user", "content": "Hello"}],
)
Python (After)
from openai import OpenAI
client = OpenAI(
base_url="https://openrouter.ai/api/v1", # Changed
api_key=os.environ["OPENROUTER_API_KEY"], # Changed
default_headers={
"HTTP-Referer": "https://your-app.com", # Added (optional)
"X-Title": "Your App", # Added (optional)
},
)
response = client.chat.completions.create(
model="openai/gpt-4o", # Prefix with provider namespace
messages=[{"role": "user", "content": "Hello"}],
)
TypeScript (After)
import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://openrouter.ai/api/v1",
apiKey: process.env.OPENROUTER_API_KEY,
defaultHeaders: { "HTTP-Referer": "https://your-app.com", "X-Title": "Your App" },
});
const res = await client.chat.completions.create({
model: "openai/gpt-4o",
messages: [{ role: "user", content: "Hello" }],
});
Model ID Mapping
| OpenAI Direct | OpenRouter ID |
|---|---|
gpt-4o |
openai/gpt-4o |
gpt-4o-mini |
openai/gpt-4o-mini |
gpt-4-turbo |
openai/gpt-4-turbo |
o1 |
openai/o1 |
o1-mini |
openai/o1-mini |
You also gain access to non-OpenAI models through the same SDK:
# Same client, any provider
response = client.chat.completions.create(
model="anthropic/claude-3.5-sonnet", # Anthropic
messages=[{"role": "user", "content": "Hello"}],
)
response = client.chat.completions.create(
model="google/gemini-2.0-flash", # Google
messages=[{"role": "user", "content": "Hello"}],
)
What Works Identically
| Feature | Status | Notes |
|---|---|---|
chat.completions.create |
Fully supported | Main endpoint, all parameters |
stream: true |
Fully supported | SSE format identical to OpenAI |
tools / tool_choice |
Supported | OpenRouter transforms for non-OpenAI providers |
responseformat: { type: "jsonobject" } |
Supported | Basic JSON mode |
responseformat: { type: "jsonschema" } |
Supported | Strict schema mode |
temperature, topp, maxtokens |
Supported | Standard parameters |
stop sequences |
Supported | Array of stop strings |
n (multiple completions) |
Supported | Multiple choices |
What Differs
| Feature | Difference | Workaround |
|---|---|---|
| Model IDs | Prefixed with provider/ |
Update model strings |
organization param |
Not used | Remove from client init |
| Embeddings | Limited support | Use direct provider or dedicated embedding service |
| Fine-tuned models | Not directly accessible | Use provider's fine-tuned model ID if hosted |
logprobs |
Model-dependent | Check model capabilities via /api/v1/models |
| Responses API | Beta support | Use /api/v1/responses endpoint |
OpenRouter-Only Features
These are available through the same SDK but are unique to OpenRouter:
# Model fallbacks (try models in order)
response = client.chat.completions.create(
model="anthropic/claude-3.5-sonnet",
messages=[{"role": "user", "content": "Hello"}],
extra_body={
"models": [
"anthropic/claude-3.5-sonnet",
"openai/gpt-4o",
"google/gemini-2.0-flash",
],
"route": "fallback",
},
)
# Provider preferences
response = client.chat.completions.create(
model="anthropic/claude-3.5-sonnet",
messages=[{"role": "user", "content": "Hello"}],
extra_body={
"provider": {
"order": ["anthropic"], # Prefer Anthropic direct
"allow_fallbacks": True,
"sort": "price", # Cheapest first
},
},
)
# Plugins (web search, response healing)
response = client.chat.completions.create(
model="openai/gpt-4o",
messages=[{"role": "user", "content": "What happened today?"}],
extra_body={
"plugins": [{"id": "web"}], # Enable real-time web search
},
)
Dual-Provider Pattern
import os
from openai import OpenAI
def create_client(provider: str = "openrouter") -> OpenAI:
if provider == "openai":
return OpenAI(api_key=os.environ["OPENAI_API_KEY"])
return OpenAI(
base_url="https://openrouter.ai/api/v1",
api_key=os.environ["OPENROUTER_API_KEY"],
default_headers={"HTTP-Referer": "https://your-app.com"},
)
# Switch providers without changing application code
client = create_client(os.environ.get("LLM_PROVIDER", "openrouter"))
Error Handling
| Issue | Cause | Fix |
|---|---|---|
| 400 unsupported parameter | Model doesn't support a parameter | Conditionally set params based on model capabilities |
| Different response quality | Non-OpenAI model handles prompt differently | Adjust prompts per model family; test before switching |
Missing organization |
OpenRouter ignores org-level auth | Remove organization from client init |
Enterprise Considerations
- Use environment variables to switch between direct OpenAI and OpenRouter without code changes
- Test your full prompt suite across providers before migrating production traffic
- Monitor response quality and latency after migration; some prompts may need tuning
- OpenRouter normalizes the API across providers, but subtle behavioral differences exist between model families
- Use
extra_bodyfor OpenRouter-specific features (provider preferences, plugins, fallbacks)