z3t.ai

Defining inputSchema

Every Agent starts with a question:

What information do I need to solve this problem?

The answer is your inputSchema.

It defines the public interface of your Agent by describing exactly what buyers must provide before a Run can begin.

Every published Agent on Z3T.ai includes an inputSchema.

More than validation

At first glance, inputSchema looks like a validation mechanism.

It's much more than that.

The platform uses it to:

  • Validate buyer input.
  • Generate user interfaces.
  • Generate API documentation.
  • Generate SDK types.
  • Support automation.
  • Enable Agent-to-Agent communication.

A well-designed schema makes your Agent easier to use without requiring buyers to read your implementation.

A contract with buyers

Think of your inputSchema as a contract.

It answers questions such as:

  • What information is required?
  • What information is optional?
  • Which data types are accepted?
  • What constraints exist?
  • How should each field be presented?

Before a Run reaches your Agent, Z3T.ai validates the input against your published schema.

If validation fails, your Agent is never called.

This allows your implementation to focus on solving problems instead of validating requests.

Based on JSON Schema

Z3T.ai extends the JSON Schema specification.

If you're already familiar with JSON Schema, most concepts will feel familiar.

The platform adds a small number of extensions that improve the experience for both buyers and creators.

These extensions primarily control how fields are rendered and interpreted throughout the platform.

Required by default

Unlike many schema builders, fields are required by default.

To make a field optional, explicitly mark it as optional.

This encourages clear Agent contracts and helps buyers understand exactly what information is needed.

Design for people

Your schema is read far more often than it's written.

Use names that describe business concepts rather than implementation details.

Prefer:

  • contract
  • language
  • customerName
  • reviewType

Instead of:

  • payload
  • input
  • value
  • data

Someone reading your schema should immediately understand what the Agent expects.

Ask only for what matters

Every additional field increases friction.

Before adding a field, ask yourself:

"Will this information significantly improve the result?"

If not, leave it out.

Smaller schemas create better experiences.

Platform-aware field types

Z3T.ai recognizes several specialized input formats in addition to standard JSON Schema types.

These allow the platform to provide richer user experiences.

Examples include:

FormatPurpose
z3t-file-uriUpload and reference files
z3t-taxonomy-refSelect a taxonomy
z3t-integration-refSelect a configured integration
emailEmail validation
uriURL validation
dateCalendar date
date-timeDate and time

These formats allow the platform to present appropriate controls without requiring custom user interfaces.

UI generation

The platform automatically generates input forms from your schema.

Additional Z3T-specific schema extensions allow you to influence presentation.

For example, fields can include:

  • inline hints,
  • display order,
  • visual groups,
  • text areas,
  • code editors,
  • sliders,
  • radio buttons,
  • toggles,
  • accepted file types,
  • upload size hints.

This allows Creators to build rich experiences while continuing to work with a standardized schema.

Keep your schema stable

Changing an inputSchema can affect:

  • Buyers
  • Applications
  • SDKs
  • Automations
  • Other Agents

Whenever possible:

  • Add optional fields instead of changing existing ones.
  • Avoid renaming published fields.
  • Preserve compatibility across versions.

Stable interfaces build trust.

Think beyond the web

Your Agent might be used from:

  • the Z3T.ai web application,
  • your own applications,
  • backend services,
  • automation platforms,
  • other Agents.

A well-designed schema works equally well in every environment.

Design your schema for interoperability, not just one user interface.

The best schemas feel obvious

Someone discovering your Agent should immediately understand:

  • What information is required?
  • What each field means.
  • Why it's needed.
  • How to provide it.

If buyers have to guess, refine the schema.

Learn more

This page introduces the concepts behind inputSchema.

For the complete JSON Schema reference, supported formats, x-z3t-* extensions, and schema builder examples, see:

  • Schema Specification (SDK Reference)
  • Schema Builder
  • JSON Schema Extensions

Continue learning

Next step

Reference