Your first call in five minutes, with a test key
Qomni documentation
AI agents that answer the phone, call your customers and talk in real time from your app. One API, with the same rules on every endpoint.
Start here
Tutorials: from zero to a working product
End-to-end solutions, step by step, with the time each one takes. Follow one and you end with something running.
Pick a product and make your first call
A test key is all you need: the same authentication, the same errors and the same test mode in every product.
AI agents that answer and place calls
Create an agent, give it a voice and put it to work: REST for the calls and a WebSocket to talk to it in real time from your app.
curl https://api.qomni.ai/agents/v1/calls \
-H "Authorization: Bearer sk_agents_test_…"GuidesAPI referenceapi.qomni.ai/agents/v1
Changelog
The latest changes to the APIs and to these docs.
Beta: the website, the waitlist and this documentation are public
Qomni Agents enters private beta. From today qomni.ai and this documentation are public, and you can join the waitlist. The API and the app are not open yet.
The API on a single host
The Qomni API now lives at api.qomni.ai, with the product and the version in the path: https://api.qomni.ai/agents/v1. The previous host (agents-api.qomni.ai) no longer serves requests: it responds 404 api_host_moved with the new base URL, without redirecting.
Tutorials, and a changelog per product with RSS
A tutorials section with end-to-end solutions, and this page with a product filter, a page per entry and RSS.
BYOC, connect your own telephony
Bring your carrier or your SIP trunk and your usual numbers reach your voice agents, without porting anything and without changing provider.
Blog subscriptions by email
The blog emails you about each new article: leave your address at the foot of any article or of the list, confirm from the email and you are done.
Engineering notes
How the platform is built: the decisions and why.
Products
What you can build today, each with its guide, its reference and its events.
Learn it once, use it everywhere
How a request authenticates, fails and retries is the same in every product: what you learn with one holds for the next.
One secret key per product
Sent as a bearer token; test keys start with sk_<product>_test_.
Create your keyOne error format across every product
A stable code for your code, a message for people and a request_id for support.
Handle errorsThe real API, without real consequences
The same API and validation with a test key: nothing is sent and nothing is billed.
Test safelyRetry without creating twice
Send an Idempotency-Key and repeat the request with peace of mind.
Retry safelyHow many requests you can make
120 per minute per key and product; above that, a 429 with Retry-After.
See the limitsCloudEvents with the same envelope
Every platform event arrives as a CloudEvent, with the same envelope and the same id to deduplicate on.
Receive eventsReady for your AI agent
Paste it into Claude, Cursor, Copilot or whichever agent you use: it reads this documentation and helps you integrate the APIs.
All the documentation, ready for your agent
- An index of the docs for your agentEvery page and its summary in one file an LLM reads at once.
- All the docs in one fileThe complete guides and reference, ready to hand over as context.
- Every page in Markdown tooAdd index.md to any URL, or copy the page from its actions menu.
Follow Qomni
What we publish and where to follow it.
Blog
Read the latest
Engineering notes on how the platform is built, by the people building it.
Changelog
What changed
Every release of the APIs and this documentation: what is new, changed and fixed.