1
Sign up
Create an account at agentheya.com and log in to the dashboard. You’ll land on the agent list — your workspace.
2
Create an agent
Click Create Agent. Give it a name — this becomes part of its public URL. For example, if you name it
billing-helper, it will live at agentheya.com/agent/billing-helper.You can rename it anytime from the Details page.3
Tell it what to do
Open the AI Helper tab. Describe your agent in plain language — what it does, who it’s for, and how it should behave:The AI Helper writes the purpose, rules, and behaviour guidelines for you. Review the proposal and click Apply to save it.You can refine it by continuing the conversation: “Make the tone more casual” or “Add a rule that it should never discuss competitor pricing.”
4
Enable tools
Go to the Tools tab. Toggle on the tools your agent needs.Common choices:
- Web search — lets the agent look up current information
- Collect form — structured intake (name, email, issue type)
- Send email — sends confirmation or follow-up emails
- Human handoff — escalates to you or your team
5
Add knowledge (optional)
Go to Data → Public and upload files the agent should know about: your help docs, FAQ, pricing page, product specs. The agent searches them automatically when a user asks a relevant question.Supported: PDF, Markdown, plain text, CSV, JSON, DOCX.
6
Preview and test
Click Preview in the sidebar to open a sandboxed chat. Talk to your agent as if you were a user. Try edge cases:
- Ask something it should know (from your uploaded docs or purpose)
- Ask something it should decline (check your rules)
- Trigger a tool (ask “what’s the weather?” if you enabled web search)
7
Share it
When you’re happy with the behaviour:
- Go to Details and set visibility to Public.
- Share the agent’s web chat URL:
https://agentheya.com/agent/{your-agent-name}.
- Web chat — send the URL directly, paste it in an email, or add a button to your website
- Embedded widget — add a JavaScript snippet to your website to embed a chat widget, or embed the URL in an
<iframe>(docs) - OpenAI-compatible API — call the agent from any app using the standard OpenAI SDK, just pointing at a different base URL (docs)
- MCP server — expose the agent as a tool in Claude Desktop, Cursor, Zed, or any MCP-capable host (docs)
- Schedules — run the agent automatically on a timer without anyone sending a message (docs)
What’s next
AI Helper
Use conversation to configure every part of your agent.
Configuration
Understand every setting and what it controls.
Tools
Add custom tools that connect your agent to any API.
Deploying your agent
Chat link, REST API, MCP server, and scheduled runs.