Agents

An agent is a specialized AI colleague inside Stoa. Each one is tailored to a specific job: a Compliance Agent knows your compliance documents and the right tone for an audit memo, a Grant Writing Agent knows your funder templates and your past proposals, a Clinical Guidelines Agent knows your protocols and how a clinician asks a question.

You pick an agent at the top of any chat. The right agent gives you better answers than a generic chatbot because it has been pointed at the right knowledge, given the right instructions, and limited to the right tools.

What an agent has

Every agent is configured with a few things:

• Instructions. A persona and a set of rules. "You are a senior compliance officer. Cite policy section numbers. Flag anything that needs legal review."
• Knowledge bases. The knowledge bases the agent can search. A Legal Precedent Agent might have access to your case archive but not your HR handbook.
• Skills. Optional skills it references: reusable bundles of instructions, knowledge, and tools that get folded in when they're relevant, without you having to pick them yourself.
• Tools. Optional tools the agent can call, like a customer database lookup or a records system query.
• Permissions. Who in your organization is allowed to use the agent, controlled by your administrators.

You don't have to think about any of this when you use an agent. You just pick one and start typing.

Picking the right agent

The agent picker is at the top of the chat input. The default agent is fine for general questions. For domain work, switch to the matching specialist:

• Compliance Agent for audit prep, policy interpretation, regulator-style questions.
• Grant Writing Agent for proposal drafts, donor reports, theory of change writing.
• HR Policy Agent for handbook lookups, leave policies, conduct rules.
• Clinical Guidelines Agent for protocol questions, dosing references, care pathway lookups.
• Legal Precedent Agent for case research, contract clause comparisons, redline assistance.

Your organization may have more, fewer, or different agents than this list. Hover over an agent in the picker to read its short description and see what it specializes in.

Switching agents mid-conversation

You can change agents partway through a chat. Open the agent picker and pick another one. The conversation history stays with you, but the new agent will answer your next question with its own knowledge bases, instructions, and tools.

This is useful when a question crosses departments. You start with the HR Policy Agent on a leave question, realize there's a payroll piece, and switch to the Finance Agent to finish it without losing context.

When you can't find the right agent

If none of the agents in your picker fit the work you're doing, talk to your administrator. Agents are configured centrally, so an admin can spin up a new one (with the right knowledge bases, the right instructions, and the right access) in minutes. Bring them a clear ask: "I need an agent that answers procurement questions from the procurement manual and the supplier database."

Why specialized agents beat generic chatbots

A generic chatbot has to guess what you want. A Stoa agent knows. It has been told who it's serving, what corpus to search, what tone to use, and what tools it's allowed to call. That means:

• Fewer wrong answers from unrelated documents.
• Citations that point to the documents your team actually trusts.
• A consistent voice across the team using the same agent.
• Tools that match the job, not a generic toolkit.

Pick the right agent and the rest of Stoa works the way you'd expect.

Where to next

• Chats for how to start a conversation.
• Skills for the reusable capability bundles agents can draw on.
• Tools for the systems agents can call.
• Knowledge bases for what agents search.