What Are AI Agents in Chase Agents?
Chase Agents supports two distinct modes of AI-powered work: Automations and AI Agents. While Automations execute predefined multi-step workflows autonomously, AI Agents are conversational interfaces that let you interact with your connected services through natural language in real time.
What Is an AI Agent?
An AI Agent in Chase Agents is a persistent assistant configured with a system prompt and a set of MCP server connections. When you open a chat session with an agent, the underlying language model has access to all the tools exposed by those servers. You ask questions or give instructions in plain language, and the agent decides which tools to call, executes them, and returns results within the conversation.
Agents are stateful within a session: they remember context from earlier in the conversation and can chain multiple tool calls together to answer a single question. For example, you might ask an agent to pull last week's sales data from your database, compare it to the same period last year, and draft a summary email, all in one conversational exchange.
Agents vs. Automations
The key distinction is interactivity. Automations run on a schedule or trigger and execute without human input. Agents wait for you to ask something, respond, and continue the conversation. This makes agents ideal for exploratory or ad-hoc work where the next step depends on what you learn from the previous one.
Use an Automation when you know the exact steps and want them to run repeatedly without intervention. Use an Agent when you want to have a dialogue with your data and services, asking follow-up questions, drilling into details, or making decisions based on live information.
Agent Architecture
Every agent is defined by three things:
- Title: the display name for the agent within your workspace.
- System Prompt: instructions that shape the agent's behavior, persona, focus area, and response style.
- Server Mappings: the set of MCP connections and their exposed tools that the agent is allowed to use.
When a user sends a message, the agent receives the conversation history, its system prompt, and the full list of available tools. It reasons about which tools to call, executes them against your live connections, incorporates the results, and generates a response. This loop repeats until the agent determines it has fully answered the user's request.
Agents and the Model Context Protocol
Chase Agents exposes its full tool engine to external AI clients through MCP. This means you can connect Claude Desktop, Cursor, or any other MCP-compatible AI assistant to your Chase Agents workspace and interact with your agents and automations directly from within those tools. The platform acts as both an agent runtime and an MCP server, making it composable with the broader AI ecosystem.
When Agents Shine
AI Agents are particularly well-suited for exploratory data analysis where you do not know the shape of the answer in advance, drafting and reviewing content that requires judgment, answering one-off questions about live data without writing any code, and workflows where the next step depends on what the previous step returned. For repetitive, scheduled, or fully autonomous work, Automations remain the better choice.