Creating an Agent

Create autonomous agents by selecting types, modes, and assigning tools and documents

Overview

Agents in OpenRails are autonomous entities that can perform tasks using tools, access documents, and interact with external services. Each agent has a type that determines its role in workflows: producer, transformer, or writer.

Agents and Symphonies overview

Agent Types

Type Description Typical Use
Producer Generates content, data, or analysis from scratch Research, content generation, data extraction
Transformer Takes input and transforms it into a different format or structure Data transformation, translation, summarization
Writer Produces final output such as reports, documents, or database entries Report generation, database writes, file output

Create an Agent

Navigate to Agents

From the sidebar, click Agents to view the agent list for your project.

Click "New Agent"

Click New Agent to open the agent creation form.

Enter Agent Details

Fill in the basic configuration:

  • Agent Name — A descriptive name for the agent
  • Description — What this agent does and when to use it
  • Type — Select Producer, Transformer, or Writer
  • LLM Model — Choose the model that powers this agent

Write the System Prompt

Define the agent's behavior with a detailed system prompt. Include instructions for how the agent should use its tools, what outputs to produce, and any constraints on its behavior.

Assign Tools

Select the tools the agent can use from the tool catalog. See Configuring Agent Tools for details on available tools.

Link Documents

Optionally link data lakes to give the agent RAG access to your documents during execution.

Set Budget Limits

Configure execution constraints:

  • Time Limit — Maximum execution time per run
  • Step Limit — Maximum number of tool invocations per run

Save the Agent

Click Save to create the agent. You can now run it manually or include it in symphonies and playbooks.

Tip: Start with conservative budget limits and increase them as you gain confidence in the agent's behavior. This prevents runaway executions during development.
Important: Agents with database or web tools can modify external systems. Always test new agents in a staging environment before deploying to production.

Next Steps