Free Chatbot API: Free Tiers Compared, Plus a Free Key
Looking for a free chatbot API for a bot, an app or a small agent? Below we compare the free tiers from Google Gemini, Groq, Mistral, OpenRouter and Hugging Face, using only what each provider publishes. Every SentioGate account also gets an OpenAI-compatible key on the free community pool: 600 requests every 5 hours, with no credit card and no provider keys.
Which chatbot API is free?
Short answer: as of 3 Nov 2026, the Google Gemini API, Groq, Mistral's Free mode, OpenRouter's :free models and SentioGate's free community pool all let you call a chat model for free. Hugging Face Inference Providers is not free on a free account: it includes no monthly credits. Every free tier is rate-limited, and none is unlimited.
All of them accept the OpenAI chat format, so the same chatbot code runs on any of them once you change the base URL and the key. For trial-only and retired services as well, see the full free LLM API list.
Free chatbot APIs compared
Each row comes from the provider's own page, checked on 3 Nov 2026 and linked in the last column. Where a page publishes no number, we say so rather than guess. Free tiers change often, so confirm on the linked page before you rely on a limit.
| Provider | What is free | Limits (as published) | Credit card to start? | Source |
|---|---|---|---|---|
| SentioGate free community pool | Send model "auto" with no provider keys added | 600 requests per 5 hours per account; HTTP 429 with Retry-After when used up | ✓ No | Pricing |
| Google Gemini API | Free tier with limited access to certain models | Requests and tokens per minute and requests per day, per project (not per key); daily quota resets at midnight Pacific | ✓ No: billing is only needed for paid tiers | Pricing, Rate limits |
| Groq | Free plan | Per-model limits per minute and per day, published on its rate-limits page; they apply to the whole organization | ✓ No: a payment method is only needed to upgrade | Rate limits, Billing FAQs |
| Mistral | Free mode: create API keys and use the included monthly usage | Per model, in tokens per minute and requests per second; your numbers are on your Limits page | Not stated on its usage-limits page | Usage limits |
| OpenRouter free models | Model variants whose ID ends in :free | 20 requests per minute; 50 per day with under 10 credits bought, 1,000 per day with 10 or more | ✓ No, for the 50-per-day level | Limits |
| Hugging Face Inference Providers | Not free on a free account: no monthly credits are included | Free users buy credits; PRO accounts get $2 of credits a month | Credits must be bought | Pricing |
We build SentioGate, so read its row with that in mind. Two things the table can't show: Gemini and Groq apply limits per project or organization, so a second key adds no capacity, and Google says content sent on the Gemini Free tier is used to improve its products. The free LLM API list has the full notes.
Get a free chatbot API key in 3 steps
The free community pool is SentioGate's shared free tier for accounts that have not added their own provider keys yet. It speaks the OpenAI Chat Completions format, so chatbot code written for OpenAI works once you change the base URL and key.
- Step 1
Create a free account
Sign up with an email address. No credit card and no provider API keys are needed to start.
- Step 2
Create a gateway key
In the Vault, create a gateway key. It starts with sg_live_ and is shown once, so store it like a password.
- Step 3
Point your OpenAI client at SentioGate
Set the base URL to https://api.sentiogate.dev/v1 and send model "auto" to /v1/chat/completions. With no provider keys added, the free community pool answers.
| Allowance | 600 requests per rolling 5-hour window, per account |
|---|---|
| API format | OpenAI-compatible /v1/chat/completions, works with the official openai SDKs |
| Model | Send model "auto". SentioGate picks the model, and it can change |
| When busy | The pool is shared: at capacity you get HTTP 429 with a Retry-After header |
| Your own keys | Adding a provider key switches your account to BYOK routing and turns the pool off |
The pool is meant for prototypes, demos and hobby bots. For production traffic, add your own provider keys (free tiers work too) or see pricing for the upcoming paid plans.
Code example: cURL, Python or Node.js
Point any OpenAI-compatible client, or our open-source SDKs (Apache-2.0) sentiogate-python and @sentiogate/sdk, at https://api.sentiogate.dev/v1. The SentioGate hosted service itself is proprietary.
# pip install openai
from openai import OpenAI
client = OpenAI(
base_url="https://api.sentiogate.dev/v1",
api_key="sg_live_your_gateway_key", # free SentioGate gateway key
)
history = [{"role": "system", "content": "You are a friendly support chatbot."}]
def chat(user_message: str) -> str:
history.append({"role": "user", "content": user_message})
r = client.chat.completions.create(model="auto", messages=history)
reply = r.choices[0].message.content
history.append({"role": "assistant", "content": reply})
return reply
print(chat("Hi! What can you help me with?"))Open-source SDK (Apache-2.0): sentiogate-python on PyPI. The SentioGate service itself is proprietary.
Free chatbot API in Python
Install the official openai package and set the base URL. Free tiers return HTTP 429 when you hit a limit. The library already waits and retries a 429 a few times (max_retries); if the limit is still in force after that, catch RateLimitError and read the Retry-After header so your bot can tell the user when to come back.
# pip install openai
import os
from openai import OpenAI, RateLimitError
client = OpenAI(
base_url="https://api.sentiogate.dev/v1",
api_key=os.environ["SENTIOGATE_API_KEY"], # sg_live_... gateway key
max_retries=3, # the SDK backs off and retries 429s for you
)
def reply(messages: list[dict]) -> str:
try:
r = client.chat.completions.create(model="auto", messages=messages)
return r.choices[0].message.content
except RateLimitError as e:
wait = e.response.headers.get("retry-after")
return f"The free allowance is used up. Try again in {wait} seconds." if wait else "Busy, try again later."
print(reply([{"role": "user", "content": "Suggest a name for a coffee shop bot."}]))The same code works with the other free tiers in the table: change base_url, the key and the model name. For a longer chat loop with history, see the free chatbot API guide.
Add a free chatbot API to your website
A chatbot on a website needs a small backend route. Your page sends the conversation to your own server, and the server adds the API key and calls the chatbot API.
// server.js (Node 18+): npm i express
import express from 'express'
const app = express()
app.use(express.json())
app.post('/api/chat', async (req, res) => {
const r = await fetch('https://api.sentiogate.dev/v1/chat/completions', {
method: 'POST',
headers: {
Authorization: `Bearer ${process.env.SENTIOGATE_API_KEY}`, // never sent to the browser
'Content-Type': 'application/json',
},
body: JSON.stringify({ model: 'auto', messages: req.body.messages }),
})
res.status(r.status).json(await r.json())
})
app.listen(3000)// In your page: call your own route, not the API directly
async function ask(messages) {
const r = await fetch('/api/chat', {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify({ messages }),
})
if (r.status === 429) return 'Busy right now, please try again shortly.'
const data = await r.json()
return data.choices[0].message.content
}Any server works: a serverless function, a Python app or your existing API. Store the key in an environment variable and add your own per-visitor limits so one visitor can't use up the shared allowance.
Free chatbot API without a credit card or API key
- No credit card: the SentioGate free community pool, the Gemini API Free tier, Groq's Free plan and OpenRouter's
:freemodels at the 50-requests-per-day level can all be used without a card. Mistral's usage-limits page doesn't say, so check when you sign up. - No provider API key: with SentioGate you don't need an OpenAI, Gemini or Groq key. You do need a free SentioGate gateway key (it starts with
sg_live_). A request without one is rejected with HTTP 401invalid_api_key. - No API key at all: hosted APIs all ask for a key, because that is how they count your requests. The only option without one is a model on your own machine, covered next.
Open-source chatbot API options
If you want an open-source chatbot API, run an open-weight model yourself. Ollama serves an OpenAI-compatible API at http://localhost:11434/v1/. The openai client still wants an api_key value, but Ollama ignores it. Ollama can also run GGUF models from the Hugging Face Hub with ollama run hf.co/{username}/{repository}.
# Run an open-weight model on your own machine (https://ollama.com)
ollama pull llama3.2
# or any GGUF model from Hugging Face:
# ollama run hf.co/{username}/{repository}
curl http://localhost:11434/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{"model": "llama3.2", "messages": [{"role": "user", "content": "Hello!"}]}'There is no per-request fee and no hosted rate limit, but you supply the hardware, and speed depends on your machine. The SentioGate SDKs are open source (Apache-2.0); the hosted service is not.
Limits: is any free chatbot API unlimited?
No. Every hosted free tier in the table has limits, and a page that promises an unlimited free chatbot API is leaving something out. The published ones:
- SentioGate free community pool: 600 requests per 5 hours per account. The pool is shared, so you can also get a 429 when it is busy.
- OpenRouter
:freemodels: 20 requests per minute, plus 50 or 1,000 per day depending on credits bought. - Gemini, Groq and Mistral: per-minute and per-day (or per-second) limits that depend on the model; see each provider's page in the table.
- Self-hosted: no request cap, but limited by your own hardware.
When you hit a limit you get HTTP 429. Wait for the time in the Retry-After header, or send the request to another provider. The OpenRouter 429 guide, Gemini 429 guide and Groq guide cover each provider.
Use several free tiers behind one endpoint
Calling one free key directly means a 429 stops your chatbot until the limit resets. An LLM router puts several keys behind one OpenAI-compatible endpoint. Add your free Gemini, Groq, Mistral or OpenRouter keys to the SentioGate Vault and send model "auto": SentioGate automatically falls back to another model if one fails. Adding a provider key switches your account to BYOK at 0% token markup, and the free community pool is turned off for it.
Prototypes and demos
Build a chatbot, a Discord or Slack bot, or a small agent against the free community pool with any OpenAI client. The request format is the same one you use later on your own keys.
Several free-tier keys behind one endpoint
Every free tier has rate limits. Add your free Gemini, Groq, Mistral or OpenRouter keys to the Vault and send model "auto": SentioGate automatically falls back to another model if one fails, at 0% token markup.
Try it in the browser first
Chat with the free community pool in the web console before you write any code.
Check latency and 429 rates
See measured time to first token, p50/p95 latency and HTTP 429 rates across major AI providers before you choose a free tier to rely on.
Guides on free tiers and rate limits
- Free LLM API list: which providers are still free, trial-only or retired, with sources.
- OpenRouter free models and 429 errors: daily caps, Retry-After headers and fallback.
- LLM router guide: what a router does and how to choose one.
- Building a chatbot on a free API: a chat loop that keeps history and handles 429s.
- Gemini 429 RESOURCE_EXHAUSTED and Groq rate limits: why they happen and how to fail over.
- BYOK LLM gateway explained: routing your own API keys at 0% token markup.
Free Chatbot API FAQ
Common questions about free chatbot APIs, limits and moving to your own keys.
Which chatbot API is free?▼
As of 3 Nov 2026, the Google Gemini API, Groq, Mistral’s Free mode, OpenRouter’s :free models and the SentioGate free community pool let you call a chat model for free, each with its own rate limits. Hugging Face Inference Providers includes no monthly credits on a free account. The comparison table on this page links each provider’s own page.
Is there a free chatbot API that does not need a credit card?▼
Yes. Create a free SentioGate account, make a gateway key (sg_live_...) in the Vault, and call https://api.sentiogate.dev/v1/chat/completions with model "auto". Until you add your own provider keys, requests are answered by the free community pool: 600 requests per rolling 5-hour window per account, with no credit card and no provider account needed.
Is there a free chatbot API without an API key?▼
Not a hosted one: hosted APIs use the key to count your requests. With SentioGate you need no OpenAI, Gemini or Groq key, only a free SentioGate gateway key; a request without one gets HTTP 401 invalid_api_key. A model you run yourself with Ollama needs no real key, because Ollama ignores the api_key value.
Is the free AI API compatible with the OpenAI SDK?▼
Yes. It uses the OpenAI Chat Completions format, so the official openai Python and Node.js libraries, LangChain and most chatbot frameworks work by changing the base URL to https://api.sentiogate.dev/v1 and using your SentioGate gateway key. Our open-source SDKs (Apache-2.0), sentiogate-python and @sentiogate/sdk, also work.
How do I call a free chatbot API from Python?▼
Install the openai package, create OpenAI(base_url="https://api.sentiogate.dev/v1", api_key=your gateway key) and call client.chat.completions.create with model "auto". The library retries HTTP 429 a few times by itself; catch RateLimitError for the case where the limit is still in force and read the Retry-After header.
Can I add a free chatbot API to my website?▼
Yes, through a small backend route. Your page sends the conversation to your own server, and the server adds the key and calls the API. Never put an API key in browser JavaScript, because anyone can read it there and use up your allowance.
Can I use this free chatbot API for Discord or Slack bots?▼
Yes. Any bot framework or library that supports OpenAI-compatible endpoints (including discord.py, discord.js, Bolt for Slack, Botpress, Flowise or LangChain) works. Set the base URL to https://api.sentiogate.dev/v1, use your SentioGate gateway key and send model "auto".
Is there an open-source chatbot API?▼
Yes, if you host it. Ollama is open source and serves an OpenAI-compatible API at http://localhost:11434/v1/ for open-weight models you run on your own machine, including GGUF models from the Hugging Face Hub. You pay with your own hardware instead of a per-request fee.
Is any free chatbot API unlimited?▼
No. Every hosted free tier has rate limits: for example 600 requests per 5 hours on the SentioGate free community pool, and 20 requests per minute on OpenRouter’s :free models. Only a model you host yourself has no request cap, and it is limited by your hardware.
What are the limits of the free community pool?▼
Each account gets 600 requests per rolling 5-hour window. The pool is shared by all free users, so at busy times a request can get HTTP 429 with a Retry-After header telling you when to try again.
Which model does the free chatbot API use?▼
Send model "auto". The free community pool serves one model chosen by SentioGate, and the model behind it can change over time. If you need a specific model, add your own provider key and request that model by name.
What happens when I add my own API keys?▼
Your account switches to BYOK (bring your own key) routing and the free community pool turns off for it. Requests then go to your own provider keys at 0% token markup, and SentioGate automatically falls back to another model if one fails. Paid plans with no monthly request cap are coming soon.
How do I avoid 429 rate limit errors when building chatbots on free AI APIs?▼
Each free tier has tight rate limits that return HTTP 429. Add free-tier keys from several providers (Google Gemini, Groq, Mistral, OpenRouter) to the SentioGate Vault and send model "auto": SentioGate honors Retry-After and automatically falls back to another model if one fails, at 0% token markup.