Scrydon

Getting Started

Sign in to your Scrydon deployment and build a working AI agent in 10 minutes.

Build a people-research agent that takes a name and returns structured information about that person. Ten minutes.

Prefer to learn by reading complete builds? Browse the Examples — fraud intelligence, ISO review cycles, NATO maritime, and more — end-to-end. Otherwise follow the tutorial below.

Assumes Scrydon is deployed and licensed. If not, see Deployment first.

Prerequisites

  • Your Scrydon URL (e.g. https://scrydon.yourcompany.internal).
  • A user account in your org — see Identity & Provisioning.
  • At least one AI integration enabled — OpenAI, Anthropic, Azure OpenAI, Bedrock, Mistral, Ollama, or self-hosted vLLM. Org admin sets this under Settings → Integrations.
  • A search integration (Exa, Linkup, Perplexity, …) if you want the agent to look things up.

Tutorial

Open the editor → New workflow → name it people-research. The default Starter block receives input on the Chat trigger as <start.input>. Drag an Agent block onto the canvas.

Open the Agent block and set:

  • Model — any from the dropdown (populated from your enabled integrations; no hardcoded vendor list).
  • System prompt"You are a people-research agent. Given a name, use your search tools to find location, profession, education. Cite sources."
  • User prompt — drag from the Starter block; the picker inserts <start.input>.

In ToolsAdd tool → pick a search-capable tool your org has installed.

Open the Chat panel → set output binding to agent1.content → send Ada Lovelace → watch the agent call its tools.

In Response format, click the magic-wand and ask:

Create a schema named person with location, profession, and education fields.

The platform generates a JSON Schema and binds it to the agent's output. Re-run from Chat.

Next steps

On this page

On this page