> ## Documentation Index
> Fetch the complete documentation index at: https://docs.usetuner.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Direct API
> Build the call timeline yourself and send it to Tuner's Create Call API — for stacks that aren't Python.
On a Python stack? Use the [Python SDKs](/docs/integrations/custom-stack/sdk-quickstart) instead — they capture the timeline for you and send it in one call. This guide is for every other language.
The API is what the SDKs call under the hood. Going direct means you own the part the observers would have handled: assembling the structured conversation timeline, and deriving per-turn timing and interruptions from whatever your stack exposes.
Concretely, you'll be building:
* the turn-by-turn timeline — user speech, agent speech, tool calls, tool results — each with start and end timestamps relative to call start
* per-turn latency fields, if you want them: STT time-to-transcript, LLM time-to-first-token, TTS time-to-first-byte, end-to-end
* interruption detection, by comparing when a user turn starts against when the agent's turn ended
For a worked example of how much that is in practice, see [Connecting to Vapi (Transient)](/docs/integrations/vapi-transient-assistants) — a single-file integration that does exactly this against Vapi's payload and call log.
***
## Prerequisites
* Tuner Active Account
* A Tuner agent with provider **Custom API**
* Your Tuner API key and workspace ID — see [the contract and authentication](/docs/integrations/custom-stack/overview#the-contract)
***
## Happy path
The video below walks through the full flow.
1. **Create a Tuner Agent** and select **Custom API** as the integration type.
2. In your voice system, after the call ends, capture:
* the **full structured conversation timeline** (agent and user messages, tool calls, and any node or state transitions)
* call start and end timestamps
* a publicly accessible recording URL
* any relevant operational or enrichment data — cost, call outcome, dynamic variables, or provider-native analysis
3. Call the **Create Call** endpoint with all data, following the API reference.
4. Open **Tuner → Calls** to see the processed results.
***
## API reference
The full **Create Call** request and response schema, with examples:
→ [Create Call API Reference](/docs/api-reference/create-call)
***
## Best practices
* **Make retries safe.** Treat `call_id` as immutable and retry the same request on failure. A duplicate `call_id` returns `409`, which counts as success.
* **Send the final transcript.** If you only have partial transcripts, send only when final, or the analysis runs on incomplete data.
* **Log request IDs and responses.** You'll want them when debugging ingestion.
***
## Common issues
### Calls don't appear in Tuner
Usually one of:
* wrong `workspace_id`
* `agent_remote_identifier` doesn't match the agent mapping you intended
* the call was sent but the transcript is empty or invalid
### Unauthorized (401)
* Ensure you're sending `Authorization: Bearer `, not `X-API-Key`.
* Confirm the key belongs to the workspace you're sending to.
### Validation errors (4xx)
* Follow the API reference for required fields and types.
* If you're unsure which field failed, start from the minimal example in the API reference and add fields incrementally.
***
## Next steps
Required fields, types, and a minimal example.
A full single-file integration built on this API.
Receive events when analysis finishes or an alert fires.
What Tuner does with the data once it lands.