AI agents that answer—and act
Each team can run an active agent on the live chat widget: grounded in published helpdesk articles, steered by personality and instructions, and optionally equipped to create drafts (posts, featured images, knowledge-base articles). Custom HTTP and LLM tasks plug into automations; everything else is available on the Hyax REST API with your team key.
Manual support doesn't scale
Your customers expect instant answers. But you can't be online 24/7, and scripted bots fall apart.
You answer the same five questions every day — and they're all in your help docs already.
Most chatbot builders require decision-tree scripting and break the moment someone asks something new.
Training a support bot means uploading docs to a third-party tool that doesn't know your products.
Off-hours mean missed messages, frustrated customers, and lost sales.
Hyax AI agents are grounded in your helpdesk articles and product data — they answer accurately, 24/7, with zero scripting.
Built-in agent capabilities
These are the toggles and behaviors you configure under App → AI Agents today.
Live chat replies
When “Respond to chats” is on and the agent is active, visitor messages on the widget are answered using your chosen model. Relevant published helpdesk articles are searched and injected into context automatically.
Models & persona
Pick from GPT-4o, GPT-5, Claude Sonnet 4.5, or Claude Haiku. Set name, avatar, personality, and free-form instructions layered on top of the system behavior.
Tool calling in chat
Optional capabilities enable multi-step tool use: create_post (draft HTML post), generate_post_image (AI featured image on a post), create_helpdesk_article (draft KB article). Without these toggles, the agent stays read-only over your help content.
Canned responses
Link saved canned replies to an agent so the model can reuse approved wording when it matches the visitor’s intent.
Custom tasks (API & LLM)
Per agent, define tasks that call your HTTP endpoints (method, URL, headers, body, optional API key) or run an LLM prompt template—with optional team brand context. Test runs are available from the dashboard; production runs are triggered from automations via the AI agent action.
REST API for everything else
Team API keys authenticate /hyax-api/v1/* routes: posts, people, email, products, orders, analytics, and more—so external agents and backends can publish or sync data without going through the widget.
API examples
Use your team key as Authorization: Bearer hx_… on your Hyax origin (see API docs for full request and response shapes).
Create a post (direct HTML)
Matches the “Create post” endpoint: send title and content as HTML. Status defaults to draft; set PUBLISHED when ready.
curl -X POST "https://platform.hyax.com/hyax-api/v1/posts" \
-H "Authorization: Bearer hx_your_team_key" \
-H "Content-Type: application/json" \
-d '{
"title": "Weekly digest: product updates",
"content": "<h2>This week</h2><p>Shipped improvements to chat and automations.</p>",
"description": "A short meta description for SEO.",
"status": "DRAFT",
"tags": ["product", "changelog"]
}'Create a post (AI generation mode)
Omit body and set generation flags; the API returns a job id. Poll GET /hyax-api/v1/posts/jobs/:id until completed.
curl -X POST "https://platform.hyax.com/hyax-api/v1/posts" \
-H "Authorization: Bearer hx_your_team_key" \
-H "Content-Type: application/json" \
-d '{
"title": "Why operators are adopting AI agents",
"generateContent": true,
"generateDescription": true,
"generateTags": true,
"generateImage": true,
"model": "gpt-5",
"status": "DRAFT"
}'Publish or update a post
After you have a post id (from create, list, or the agent’s create_post tool), patch status or body. Slugs must stay unique per team.
curl -X PATCH "https://platform.hyax.com/hyax-api/v1/posts/POST_ID" \
-H "Authorization: Bearer hx_your_team_key" \
-H "Content-Type: application/json" \
-d '{ "status": "PUBLISHED" }'Dashboard vs. HTTP API
Agent CRUD, canned-response links, and custom tasks are managed in the app via the authenticated API layer (same session as the dashboard). Integrations and headless workflows should use versioned REST under /hyax-api/v1 with a team API key—ideal for schedulers, CI, or third-party AI stacks calling Hyax alongside your widget agent.
Widget agent
One active agent per team answers chat when enabled; tools and KB search run server-side.
Automations
Invoke an agent’s custom API or LLM task from a workflow (with person and context placeholders)—see Automations.
Webhooks & events
Subscribe to commerce, audience, and content events to trigger your own agents or pipelines outside Hyax.
- →Posts — list, create (direct or AI job), get, patch, archive
- →People — profiles, tags, sequences, transactional email
- →Products & orders — catalog and order retrieval for ops bots
- →Analytics — event ingestion and reporting endpoints
Pair agents with your stack
Configure the chat agent in Hyax, call Hyax from your own code with team keys, and wire automations to custom tasks—without fictional SDKs.