Agents

An agent is a logical grouping of prompts within a project. It represents one AI-powered feature, bot, or assistant in your application.

What an Agent Is

Agents map directly to real AI agents or bots in your codebase. Some examples:

• Customer Support Bot — system prompt, escalation prompt, fallback response
• Code Reviewer — review instructions, severity guidelines, summary template
• Email Drafter — tone instructions, signature template, subject line generator
• Onboarding Assistant — welcome message, feature explanation prompts

Each agent can hold multiple prompts. This lets you version and deploy each piece of an agent's behavior independently.

Agent Limits by Plan

The number of agents per organization is capped by your plan:

Querying by Agent in the SDK

When querying prompts, you can filter by agent ID to fetch only the prompts belonging to one agent:

// Get all prompts for a specific agent
const prompts = await raison.find({ agentId: "YOUR_AGENT_ID" });

// Get a specific prompt by agent and name
const systemPrompt = await raison.findOne({
  agentId: "YOUR_AGENT_ID",
  name: "system-prompt",
});

Agent IDs are found in the dashboard by clicking on the agent name. The ID is displayed at the top of the agent detail view.

How Agents Map to Your Code

There is no SDK concept of an "agent" as a runtime object — agents are an organizational tool in the dashboard. In your code, you reference individual prompts by their ID or by { agentId, name }. The agent grouping just makes it easier to manage many prompts in the dashboard and fetch related prompts together in the SDK.