Template
All templates

Agent builder

Step-by-step coach that helps you build AI-agents in Oniroco.
coachingUse template
Description

Description

🤖 Agent Builder Assistant

Build your own AI-agent in Oniroco without technical knowledge. This assistant guides you conversationally and step-by-step from a vague idea to a fully configured, working agent.

🎯 What does this assistant do?

The Agent Builder Assistant is your personal coach for setting up AI-agents on the Oniroco platform. It asks the right questions, thinks through your use case with you, actively explores which integrations and tools are available, and ultimately delivers a ready-made implementation guide that you can follow 1-on-1 in Oniroco.

🛠️ How does it work?

The assistant guides you through 6 structured steps:

1. Clarify use case

You explain in your own words what your agent should do. The assistant asks targeted follow-up questions about the goal, target audience, and the problem being solved — until everything is crystal clear.

2. Identify tools

The assistant actively searches through all available Oniroco integrations (Outlook, Gmail, Pipedrive, …) and tool templates, and makes 2 to 4 motivated suggestions that fit your use case. It follows the principle of minimalism: no tools "just in case", static data directly in the prompt.

3. Choose trigger

Together you determine how the agent should be started:
  • Integration trigger (e.g., "new email received")
  • Webhook trigger (external systems)
  • Time-based trigger (cron/schedule)

4. Write system prompt

Through a series of targeted questions, the assistant builds a strong, structured prompt with you in a fixed 4-block structure:
  • Role & Goal
  • Step-by-step plan
  • Rules & Limitations
  • Output format

5. Recommend model

Based on complexity and volume, you get concrete advice:
  • Claude Haiku 4.5 — fast and cheap for classification/routing on high volume
  • Claude Sonnet 4.6 — the standard workhorse for most agents
  • Claude Opus 4.7 — for complex multi-step reasoning
Including the right reasoning level (Minimal / Low / Medium / High) where applicable.

6. Deliver implementation guide

You receive a complete, structured overview with:
  • ✅ Trigger configuration step by step
  • ✅ Which tools to activate + where to set approval during testing phase
  • ✅ The complete prompt in a copy-paste code block
  • ✅ Model choice + reasoning level
  • ✅ Testing instructions and fine-tuning tips
  • ✅ Explanation of duplicating for multiple users (if needed)

💡 What makes this assistant strong?

  • Truly platform-aware — doesn't invent tools, triggers or parameters, but checks live which ones are available
  • Minimalist — only adds what is strictly necessary, puts static data in the prompt
  • Safe by default — recommends approval for destructive actions during testing phase
  • Account-aware — understands that tools work per connected account and suggests the right pattern for multiple users
  • Iterative — you can always come back to ask for adjustments or share sample events

🎬 For whom?

Anyone who wants to build an AI-agent in Oniroco — whether you're a bookkeeper who wants to automate your inbox, a sales team that wants to streamline lead follow-up, or an operations manager who wants to outsource recurring workflows to AI. No technical knowledge required.

🚀 Just start with one sentence

"I want an agent that solves [problem] for [target audience]."
We'll build the rest together.