.png)
3 months ago
3
The Poe API provides access to hundreds of AI models and bots through a single OpenAI-compatible endpoint. Switch between frontier models from all major labs, open-source models, and millions of community-created bots using the same familiar interface.
Key benefits:
- Use your existing Poe subscription points with no additional setup
- Access models across all modalities: text, image, video, and audio generation
- OpenAI-compatible interface works with existing tools like Cursor, Cline, Continue, and more
- Single API key for hundreds of models instead of managing multiple provider keys
If you're already using the OpenAI libraries, you can use this API as a low-cost way to switch between calling OpenAI models and Poe hosted models/bots to compare output, cost, and scalability, without changing your existing code. If you aren't already using the OpenAI libraries, we recommend that you use our Python SDK.
Using the OpenAI SDK
- Python
- Node.js
- cURL
import os, openai
client = openai.OpenAI(
api_key=os.getenv("POE_API_KEY"),
base_url="https://api.poe.com/v1",
)
chat = client.chat.completions.create(
model="GPT-4o",
messages=[{"role": "user", "content": "Top 3 things to do in NYC?"}],
)
print(chat.choices[0].message.content)
Options
-
Poe Python Library (✅ recommended)
Install with pip install fastapi-poe for a native Python interface, better error handling, and ongoing feature support.
→ See the External API Guide to get started.
-
OpenAI-Compatible API (for compatibility use cases only)
Poe also supports the /v1/chat/completions format if you're migrating from OpenAI or need a REST-only setup.
Base URL: https://api.poe.com/v1
For new projects, use the Python SDK—it’s the most reliable and flexible way to build on Poe.
Known Issues & Limitations
Bot Availability
- Private bots are not currently supported - Only public bots can be accessed through the API
- The Assistant bot is not available via the OpenAI-compatible API endpoint
Media Bot Recommendations
- Image, video, and audio bots should be called with stream=False for optimal performance and reliability
Parameter Handling
- Best-effort parameter passing - We make our best attempts to pass down parameters where possible, but some model-specific parameters may not be fully supported across all bots
Additional Considerations
- Some community bots may have varying response formats or capabilities compared to standard language models
API behavior
Here are the most substantial differences from using OpenAI:
- The strict parameter for function calling is ignored, which means the tool use JSON is not guaranteed to follow the supplied schema.
- Audio input is not supported; it will simply be ignored and stripped from input
- Most unsupported fields are silently ignored rather than producing errors. These are all documented below.
Detailed OpenAI Compatible API Support
Request fields
| model | Use poe bot names |
| max_tokens | Fully supported |
| max_completion_tokens | Fully supported |
| stream | Fully supported |
| stream_options | Fully supported |
| top_p | Fully supported |
| tools | Not Supported |
| tool_choice | Not Supported |
| parallel_tool_calls | Not Supported |
| stop | All non-whitespace stop sequences work |
| temperature | Between 0 and 1 (inclusive). Values greater than 1 are capped at 1. |
| n | Must be exactly 1 |
| logprobs | Ignored |
| store | Ignored |
| metadata | Ignored |
| response_format | Ignored |
| prediction | Ignored |
| presence_penalty | Ignored |
| frequency_penalty | Ignored |
| seed | Ignored |
| service_tier | Ignored |
| audio | Ignored |
| logit_bias | Ignored |
| store | Ignored |
| user | Ignored |
| modalities | Ignored |
| top_logprobs | Ignored |
| Reasoning_effort | Ignored |
Response fields
| id | Fully supported |
| choices[] | Will always have a length of 1 |
| choices[].finish_reason | Fully supported |
| choices[].index | Fully supported |
| choices[].message.role | Fully supported |
| choices[].message.content | Fully supported |
| choices[].message.tool_calls | Fully supported |
| object | Fully supported |
| created | Fully supported |
| model | Fully supported |
| finish_reason | Fully supported |
| content | Fully supported |
| usage.completion_tokens | Fully supported |
| usage.prompt_tokens | Fully supported |
| usage.total_tokens | Fully supported |
| usage.completion_tokens_details | Always empty |
| usage.prompt_tokens_details | Always empty |
| choices[].message.refusal | Always empty |
| choices[].message.audio | Always empty |
| logprobs | Always empty |
| service_tier | Always empty |
| system_fingerprint | Always empty |
Error message compatibility
The compatibility layer maintains consistent error formats with the OpenAI API. However, the detailed error messages may not be equivalent. We recommend only using the error messages for logging and debugging.
All errors return:
"error": {
"code": 401,
"type": "authentication_error",
"message": "Invalid API key",
"metadata": {...}
}
}
| 400 | invalid_request_error | malformed JSON, missing fields |
| 401 | authentication_error | bad/expired key |
| 402 | insufficient_credits | balance ≤ 0 |
| 403 | moderation_error | content flagged |
| 404 | not_found_error | wrong endpoint / model |
| 408 | timeout_error | model didn’t start in a reasonable time |
| 413 | request_too_large | tokens > context window |
| 429 | rate_limit_error | rpm/tpm cap hit |
| 502 | upstream_error | model backend not working |
| 529 | overloaded_error | transient traffic spike |
Retry tips
- Respect Retry-After header on 429/503.
- Exponential back‑off (starting at 250 ms) plus jitter works well.
- Idempotency: resubmit the exact same payload to safely retry.
While the OpenAI SDK automatically manages headers, here is the complete list of headers supported by Poe’s API for developers who need to work with them directly.
Response Headers:
| openai-organization | OpenAI org | Unsupported |
| openai-processing-ms | Time taken processing your API request | Supported |
| openai-version | REST API version (2020-10-01) | Supported |
| x-request-id | Unique identifier for this API request (troubleshooting) | Supported |
Rate Limit Headers
Our rate limit is 500 rpm, but theres no planned support for any of the following rate limit headers at this time:
- x-ratelimit-limit-requests (how many requests allowed for the time window)
- x-ratelimit-remaining-requests (how many requests remaining for the time window)
- x-ratelimit-reset-requests (timestamp of when the time window resets)
Getting Started
- Python
- Node.js
- cURL
import os
import openai
client = openai.OpenAI(
api_key=os.environ.get("POE_API_KEY"),
base_url="https://api.poe.com/v1",
)
completion = client.chat.completions.create(
model="gemini-2.5-pro",
messages=[{"role": "user", "content": "What are the top 3 things to do in New York?"}],
)
print(completion.choices[0].message.content)
Streaming
You can also use OpenAI's streaming capabilities to stream back your response:
- Python
- Node.js
- cURL
import os
import openai
client = openai.OpenAI(
api_key=os.environ.get("POE_API_KEY"),
base_url="https://api.poe.com/v1",
)
stream = client.chat.completions.create(
model="Claude-Sonnet-4",
messages=[
{"role": "system", "content": "You are a travel agent. Be descriptive and helpful."},
{"role": "user", "content": "Tell me about San Francisco"},
],
stream=True,
)
for chunk in stream:
print(chunk.choices[0].delta.content or "", end="", flush=True)
Migration checklist (OpenAI ➜ Poe in 60 s)
- Swap base URL – https://api.openai.com/v1 → https://api.poe.com/v1
- Replace key env var – OPENAI_API_KEY → POE_API_KEY
- Check model slugs – gpt-4o is identical, others via /v1/models.
- Delete any n > 1, audio, or parallel_tool_calls params.
- Run tests – output should match except for intentional gaps above.
Pricing & Availability
All Poe subscribers can use their existing subscription points with the API at no additional cost.
This means you can seamlessly transition between the web interface and API without worrying about separate billing structures or additional fees. Your regular monthly point allocation works exactly the same way whether you're chatting directly on Poe or accessing bots programmatically through the API.
If your Poe subscription is not enough, you can now purchase add-on points to get as much access as your application requires. Our intent in pricing these points is to charge the same amount for model access that underlying model providers charge. Any add-on points you purchase can be used with any model or bot on Poe and work across both the API and Poe chat on web, iOS, Android, Mac, and Windows.
Support
Feel free to reach out to support if you come across some unexpected behavior when using our API or have suggestions for future improvements.
