AI AGENT BUILDER

Build focused AI Agents around real capabilities

Build AI Agents with clear responsibilities, focused instructions, and only the resources they actually need. Connect structured data, documents, reusable Skills, or MCP tools, then combine specialized Agents into larger AI systems without putting every responsibility into one configuration.

Build your first Agent Read the Agent guide

Define · connect · compose

AutronAIFocused Agent

InstructionsSkillsDataDocumentsMCP tools

01Define02Connect03Compose

Responsibility first

Build an Agent around one responsibility

An effective AI Agent starts with a clear responsibility.

Instead of asking one configuration to understand every source, follow every procedure, and access every external system, AutronAI lets you create focused Agents around the kind of work they need to perform.

Each Agent has its own identity, instructions, type, and resource configuration. That separation makes it easier to understand what the Agent is responsible for, what information it can access, and which capabilities it can use.

Define

Give the Agent a name, description, type, and instructions that define its responsibility.

Connect

Select the data, documents, Skills, or tools required for that responsibility.

Compose

Attach focused Agents to a Chatbot when multiple capabilities need to work as part of one conversational system.

Behavior

Start with clear Agent instructions

Instructions define how an Agent should approach its responsibility. They should describe what the Agent is expected to do, what boundaries it should respect, and how it should use the resources connected to it.

Good instructions do not need to contain every piece of knowledge the Agent may ever use. Data, documents, Skills, and tools can remain separate resources and be connected to the Agent through the appropriate configuration.

Responsibility

State the specific job the Agent is responsible for instead of using broad instructions such as “help with anything.”

Boundaries

Define what the Agent should not attempt and what it should do when the required information or capability is unavailable.

Resource use

Explain how the Agent should use its configured source, Skill, or tool when that behavior matters.

Output behavior

Define the level of detail, structure, or constraints that are important to the task without mixing unrelated product behavior into the Agent.

Four explicit types

Choose the right Agent type

Agent type determines the kind of resource the Agent is built around. AutronAI currently separates Agents into four types so different responsibilities can remain explicit.

Explore structured data in the platform See MCP tool connections

Skills required

Custom Agent

A Custom Agent is built around reusable Skills rather than a dedicated database, document source, or MCP tool connection.

Use it when the responsibility is primarily procedural: the Agent needs clear instructions and one or more reusable Skills that describe how a task should be performed. The current builder requires at least one Skill; Custom is not an instructions-only Agent.

One source required

Structured Data Agent

A Structured Data Agent is designed for work that depends on structured information. Connect it to one configured structured source so it can work with the prepared part of a database, CSV, or Excel dataset.

The source remains independently managed, so schema access, table selection, column access, and descriptions can evolve without copying the database structure into Agent instructions.

Documents required

Document Agent

A Document Agent works with unstructured reference material. Connect one or more processed document sources for manuals, policies, guides, notes, reports, or other supported documents.

Keeping documents separate lets reference material change without redefining the Agent's fundamental responsibility.

Tools required

Tool Agent

A Tool Agent is designed for actions that depend on capabilities outside the language model. After installing an MCP server, configure the Agent with one or more specific tools it should be allowed to use.

The Agent does not automatically receive every tool from every installed server.

Clear separation

Connect resources instead of expanding the prompt

Agent instructions and Agent resources solve different problems.

Instructions tell the Agent how to behave. Resources give it the information or external capability required to perform the job. Separating the two makes the configuration easier to reason about and reduces the need to copy large amounts of changing information into instructions.

Structured Data

Provides queryable information from configured databases and tabular sources.

Documents

Provide searchable unstructured reference material.

Skills

Provide reusable instructions for specific procedures.

MCP Tools

Provide external actions and capabilities.

Instructions

Define responsibility, constraints, guidance, expected behavior, and fallback rules.

Resources

Provide queryable structured data, searchable documents, reusable Skills, and external capabilities through MCP tools.

Reusable procedure

Add reusable Skills

Skills let task-specific guidance remain reusable instead of being duplicated across Agent instructions. A Skill can describe a procedure, sequence of work, validation approach, formatting convention, or another reusable method, and remains a separate resource selected during Agent configuration.

Custom Agents

Skills are the required resource for Custom Agents in the current builder.

Other Agent types

Skills can be added alongside structured data, documents, or MCP tools when the Agent also needs reusable procedural guidance.

Why separate Skills?

When the procedure changes, the Skill can evolve independently from the Agent's identity and connected data or tools.

Least capability needed

Give Tool Agents precise MCP access

Installing an MCP server creates an external capability source for the workspace. Building a Tool Agent is where that capability becomes part of a focused AI responsibility.

AutronAI groups available tools by the installed MCP instance and lets you choose individual tools for the Agent.

Learn how AutronAI uses MCP

Choose the connection

Select from the MCP instances installed in the workspace.

Review the tools

See the tools exposed by that instance before assigning access.

Select what the Agent needs

Give the Agent only the tools relevant to its responsibility.

Combine tools with Skills

Add Skills when reusable instructions should describe how or when those capabilities are used.

The current wizard

Build in three clear steps

The builder moves from identity to resources, then gives you a final review before creation.

Basic information

Set the Agent's name, identifier, type, description, and instructions. The builder validates the Agent identifier before you continue.

Configuration

Custom → one or more Skills required.

Structured Data / SQL → structured source required; Skills optional.

Document → one or more documents required; Skills optional.

Tool → one or more MCP tools required; Skills optional.

Review

Review the Agent's identity and selected configuration before creating it. Make sure its responsibility, type, and resources agree before the capability becomes part of a larger Chatbot configuration.

Focused composition

Build small Agents instead of one oversized Agent

As capabilities grow, it can be tempting to keep adding instructions, data, documents, and tools to the same Agent.

AutronAI's Agent model makes another approach possible: keep individual Agents focused and compose multiple capabilities at the Chatbot level.

Structured data

One Agent can work with structured data.

Documents

Another can retrieve from documents.

External tools

Another can use selected external tools.

Skill-driven behavior

A Custom Agent can provide focused behavior through Skills.

Access boundary

Keep data and tool access intentional

Agent configuration is also an access boundary. A focused Agent should receive only the sources and tools necessary for its responsibility.

Structured sources

Configure the relevant tables and columns at the data-source level before connecting the source to the Agent.

Documents

Attach only the reference material relevant to the Agent's task.

MCP tools

Select individual tools instead of granting broad external capability by default.

Secrets

Credentials belong in the appropriate integration, data-source, or MCP credential fields, not inside Agent instructions.

Independent parts

Evolve an Agent without rebuilding the whole system

Agent configuration can change as the responsibility changes.

Instructions can be refined. Skills can be added or removed. Document selections can evolve. Tool access can change as MCP capabilities change.

Because these pieces are separated, changing one capability does not require rebuilding every other part of the AI system.

Composition and deployment

Move from Agent to Chatbot

An Agent is a focused capability. A Chatbot is where multiple capabilities can be brought together for interaction and deployment.

After creating the Agents you need, attach them to a Chatbot and use Chatbot Studio to review the composition alongside Memory.

From there, the Chatbot can be connected to supported communication channels or, when eligible, participate in Automation workflows.

Explore the AI Agent platform See supported channels Understand Agent Memory Explore AI Automation

Design for change

A builder for systems that can evolve

Building an AI Agent should not require deciding every future capability on day one.

Start with a focused responsibility. Give the Agent the smallest useful set of resources. Test how it behaves. Add Skills when reusable procedures emerge. Add data or documents when the task needs reliable information. Add MCP tools when the Agent needs to act outside the model.

When another responsibility becomes materially different, create another focused Agent instead of turning the original Agent into an increasingly difficult configuration.

AutronAI is designed so those capabilities can later be composed into a broader system.

FAQ

Questions, answered

Frequently asked questions

What can I build with the AutronAI AI Agent Builder?

You can create focused Custom, Structured Data / SQL, Document, and Tool Agents. Each type has different resource requirements, while instructions and selected capabilities define the responsibility the Agent should perform.

Which Agent type should I choose?

Choose a Structured Data Agent for scoped database or tabular queries, a Document Agent for searchable reference files, a Tool Agent for selected MCP actions, and a Custom Agent for focused Skill-driven behavior that does not need a dedicated data, document, or tool workflow.

Are Skills the same as MCP tools?

No. A Skill provides reusable procedural guidance about how work should be approached. An MCP tool provides an external action through a configured connection. A Tool Agent can combine instructions, Skills, and selected MCP tools when all three are useful.

Why build several focused Agents instead of one large Agent?

Focused Agents keep instructions, data access, document scope, and tool permissions easier to inspect and maintain. A Chatbot can compose several specialists when a use case needs broader coverage, without merging every responsibility into one configuration.

Can Agent resources change after creation?

Yes. The current builder supports refining instructions and changing the relevant Skills, data, documents, or MCP tool selection. Those changes let the capability evolve independently from other Agents in the Chatbot composition.

Build with AutronAI

Build your first focused AI Agent

Define its responsibility, connect the resources it needs, and expand the system as new capabilities become useful.

Start building Read the AutronAI Guide