Jonathon's AI Wiki

Voice AI & Conversation AI Public APIs

5 min read

Source: ai-research/ghl-2026-07-02-drain/support-solutions-articles-155000006379-voice-ai-public-apis.md

HighLevel exposes two adjacent, separately-documented public API surfaces for programmatically managing AI agents: Voice AI Public APIs (phone-call agents) and the Conversation AI Public API (chat/SMS/omnichannel agents). Both let a developer create, update, retrieve, and delete agents and their actions outside the in-app UI — turning what used to be a manual configuration screen into something a pipeline or another AI agent can drive. This is the answer to “how far does programmatic Voice AI / Conversation AI agent management reach” — referenced but not detailed in Voice AI Templates.

Key Takeaways

  • Two distinct APIs, two distinct products. Voice AI Public APIs manage phone-call agents (calls, transcripts, call logs). The Conversation AI Public API manages chat/SMS/omnichannel agents (actions, agents, generations). Don’t conflate them — they have separate endpoint families and separate support articles.
  • Full CRUD on agents, not just read access. Both APIs support Create, Read, Update, and Delete on the agent object itself — this is genuine programmatic agent management, not just a read-only reporting layer.
  • Actions are a first-class, separately-managed sub-resource. An agent’s “actions” (webhook invocation, appointment booking, call transfer, follow-up settings) can be attached, listed, updated, and removed independently of the agent’s core config.
  • Voice AI adds Call Log APIs. List and get call logs/transcripts, filterable by agent, contact, call type, action type, and date range (IANA timezone-aware), with 1-based pagination and sorting.
  • Conversation AI adds Generations APIs. Get Generation Details retrieves a specific AI-generated response/output for analytics, QA, or compliance export.
  • Dual authentication on Conversation AI: PIT or JWT. Personal Integration Tokens (quick, scoped, good for server-to-server) and JWT (OAuth-based app flows) are both supported — use a Sub-Account (location) token, not an Agency token, so calls act within the correct location context.
  • Webhooks close the loop. Both surfaces support webhooks for real-time call outcomes, transcripts, and summaries pushed to external systems rather than polled.

Voice AI Public APIs

Agents APIs

Create, list, read, update, and delete Voice AI agents — roll out consistent configurations across locations and environments.

EndpointWhat it does
Create AgentCreate a new Voice AI agent configuration and settings
Patch AgentPartially update an existing Voice AI agent
List AgentsRetrieve a paginated list of agents for a given location
Get AgentRetrieve detailed configuration and settings for a specific agent
Delete AgentDelete an agent and all its configurations

Actions APIs

Actions define what an agent can do during a call — invoking a webhook, retrieving data, posting results.

EndpointWhat it does
Create Agent ActionCreate a new action for a Voice AI agent
Update Agent ActionModify an existing action’s behavior/configuration
Get Agent ActionRetrieve a specific action’s configuration, including actionParameters
Delete Agent ActionPermanently remove an action and its configuration

Call Log APIs

EndpointWhat it does
List Call LogsCall logs for Voice AI agents scoped to a location. Filter by agent, contact, call type, action types, and date range (IANA timezone). Sortable, 1-based pagination.
Get Call LogReturns a single call log by call ID

Voices (Catalog APIs)

Referenced in the official table of contents as a catalog of available voices; see the live API Reference for parameters.

Conversation AI Public API

Authentication (PIT & JWT)

  1. Open Settings → Private Integrations in the HighLevel sub-account (location).
  2. Give it basic info (name and description).
  3. Select the Conversation AI scopes.
  4. Create the token and copy it.
  5. Use a Sub-Account token — Agency tokens are not the right context for these calls.

Choose PIT for simple server-to-server integrations; use JWT for OAuth app flows. Both work.

Endpoint families

FamilyEndpoints
AgentsCreate An Agent, Search Agents, Update Agent, Get Agent, Delete Agent
ActionsAttach Action To Agent, List Actions For An Agent, Get Action By ID, Update Action, Remove Action From Agent, Update Followup Settings
GenerationsGet Generation Details

FAQ highlights

  • Do I need a Sub-Account or Agency token? Sub-Account — calls need to act within the correct location context.
  • Can I manage agents through the API? Yes — create, update, retrieve, and manage programmatically.
  • Where do I find the agentId? Create or search agents via the Agents API, then use the returned id field in subsequent calls.

How this relates to AI-template Marketplace listings

Voice AI Templates covers the one-time publish-a-template-to-Marketplace flow (bundling an agent + workflows + calendar + KB for sale). These APIs are the ongoing runtime management layer — once an agent exists (whether hand-built or installed from a template), these endpoints let you read/update/delete it and its actions programmatically, and pull call logs or generations for analytics. The two are complementary: templates handle distribution, these APIs handle operation.

Try It

  • Decide which surface you need first: Voice AI (phone calls) or Conversation AI (chat/SMS) — they’re separate APIs with separate scopes.
  • Generate a Private Integration Token scoped to Conversation AI (or the equivalent Voice AI scopes) under Settings → Private Integrations in a test sub-account.
  • Call List Agents / Search Agents first to get a real agentId before attempting update/delete calls.
  • If you’re building a Marketplace app rather than a single-location integration, use JWT/OAuth instead of a PIT so the flow works across installs.
  • Wire the relevant webhook (call outcomes for Voice AI, generation/action events for Conversation AI) so you’re not polling for call transcripts or generation results.
  • Cross-check scopes needed against OAuth 2.0 before submitting a Marketplace app that uses these endpoints — request the minimum scope set.

Open Questions

  • Whether Voice AI’s “Voices” catalog API is read-only (browsing available voices) or also supports custom voice registration.
  • Rate limits specific to these two API families — not confirmed against the general API Rate Limits figures in this research pass.
  • Whether Get Generation Details (Conversation AI) exposes token/cost usage per generation, useful for cost-tracking automations.

Graph View

Backlinks