Assistant templates
A template is a curated pre-built assistant — a system prompt, model pick, voice configuration, and channel compatibility set that the agent wizard applies in one click. Templates are read-only catalog entries; the only write on this surface isinstantiate, which clones a template into a new assistant in your organization. Field-level reference lives in the Agents templates endpoints; this page walks the flow end to end.
Base path: /api/v1/agents/templates
Authentication: API key (X-API-Key) or session JWT. Reads are open to any authenticated role; instantiate requires owner, admin, or developer.
Every response below carries the full { data, meta } envelope and at least one labelled error envelope — the same floor as the SDK sample coverage policy.
Worked sequence: browse, read, instantiate, test
1. List the catalog
The catalog’stemplates array is the plain pre-rendered list; there is no server-side filter — filter by agent_type on your side when you only want, say, voice.
200:
2. Read one template detail
Before instantiating, fetch the template’s id so the call can surface its prompt fields. Unknown ids answer404 with a labelled error envelope:
404 Error
3. Instantiate the template into your tenant
POST instantiate clones the template into a new assistant row, renders the prompt against your organization’s jurisdiction, and returns the new assistant record. Repeat instantiate with a stable Idempotency-Key header to dedupe lobby-level retries.
200:
agt_a41c829e77d902) goes straight into the Assistant lifecycle — create, deploy, list, fetch, update, delete. The config.template_id back-pointer lets you trace the instance to the curated source during support.
4. Test the instantiated assistant end to end
Send one message against the new assistant id over the conversation API —POST /api/v1/agents/{id}/conversations opens a conversation you can step through turn by turn. The conversation path only exists once instantiate returned 200. An invalid id answers the same 404 NOT_FOUND envelope shown in step 2.
See also
- AI assistants — the lifecycle for the instantiated agent
- Agents API — field-level reference
- Agent model presets — the curated STT+LLM+TTS bundles for voice agents