From sign-in to a scheduled, knowledge-grounded agent. This walk-through builds an agent that answers support requests from your own documentation. It uses only built-in features and one model credential.
Steps
1. Sign in and create a project
2. Add a model credential
3. Index documents
4. Build the workflow
5. Schedule or trigger it
Before anyone relies on the agent, add a Tests entry for the model credential and request a review to promote the workflow to production.
- Sign in. A first installation has a one-time bootstrap administrator whose password you change at first sign-in.
- Open Projects › Create Project (or press ⌘K / Ctrl+K and choose Create project) and name it.
- Open Manage › Credentials in the project and click New Credential.
- Choose the provider type — Azure OpenAI, Azure AI Foundry, Gemini, Vertex AI, OpenAI or Anthropic — and fill in key, endpoint or deployment.
- Click Test, then save.
- Open Build › Knowledge Base, create a knowledge base and set its embedding connection in settings.
- Click Upload & Embed and add a few PDFs or Word files. Wait until every document shows as indexed.
- Open Build › Workflows › Create from Template and choose User Request AI Solver.
- Open KB Search and select your knowledge base. Open AI Solver and select your model credential.
- Click Run, type a question, and open the execution log to see each step.
- To run it on a timetable, open Scheduling & Runs, create a schedule, pick the workflow, frequency and time zone, and check the preview.
- To run it on events, replace the Input node with a Webhook node and paste its address into the source system.
This site needs JavaScript for the full experience. Every page above is reachable without it.