Skip to main content
Clarity voice agents answer and place phone calls for you. An agent listens to the caller, reasons with a language model, uses your tools (look up an order, book a slot, transfer to a person), and speaks back. Every call is recorded, transcribed, and kept as a conversation you can read back from the API.

​
Where calls come from

You set up numbers, routing and agents in Clarity at support.onclarity.com. The API covers the parts teams usually automate.

​
The model

1

A project holds your agents

Every path in the Voice API starts with a project: /v1/projects/{projectId}/…. List the projects your key can reach with GET /v1/projects.
2

An agent has versions

Only the published version takes live calls. Edits go to a draft, which you publish when it is ready. Older published versions stay in the history and can be republished. See Manage agents.
3

A call is a conversation

Each call creates a conversation. A conversation can have several call legs (calls), for example when a call is transferred.
4

Conversations carry the record

After the call ends you can read its transcript, a signed link to the recording, the tool calls the agent made, the summary and any fields the agent extracted. See Conversations and recordings.

​
Base URL and authentication

Every request uses your organization API key in the Authorization header. The key reaches every project of your organization. See Authentication. Most errors come back as JSON with an error field. Some errors add a machine-readable code, for example AGENT_NOT_PUBLISHED.

​
What you can automate

Place outbound calls

Have an agent call a customer and get the conversation id back right away.

Get notified when a call ends

Receive the call’s transcript, parties and extracted fields as a vCon document at your HTTPS endpoint.

Read conversations and recordings

List calls, filter them, and pull transcripts, recordings and metrics.

Run campaigns

Call a list of numbers with per-contact variables, retries and calling hours.

Manage agents

Create and edit agents, publish drafts, and roll back to an earlier version.

Test with evals

Run a suite of scripted test calls against an agent version and read the judge’s verdicts.
The full list of endpoints, with request and response schemas, is in the API reference.