Skip to content

Agent Catalogue

The Agent Catalogue is a single place to organize, package, and publish agents as reusable units. Each catalogue entry is an Agent Card: a bundle of everything needed to understand, use, and interact with that agent.

Overview

With the catalogue you can:

  • Discover available agents
  • Package existing agents into reusable cards
  • Upload agents as ZIP packages to import prebuilt definitions
  • Download agents as ZIP packages for sharing, backup, or migration
  • Attach supporting assets such as workflows, personas, and documentation
  • Share standardized agent definitions across teams

The Agent Card model aligns with the A2A Agent Discovery approach: each agent is represented as a structured, discoverable entity.

Browse catalogues on the web

Explore published agent catalogues in your browser:

Fabrix.ai Agent Catalog

1. Agent Card

An Agent Card is a packaged view of an agent. Each card includes:

  • Agent metadata — Identity, description, versioning, and classification
  • Associated workflows and personas — How the agent runs and behaves
  • Supporting assets — Documents, images, videos, and other files
  • Interaction and observability entry points — How to run the agent and inspect its behavior

Together, these elements describe what the agent does and how to use it.

Agent card view

An agent card shows metadata (version, publisher, category, status), description, primary actions (Interact, View agentic workflows, Observe), and tabs for documents, images, and contact information.

Agent card: Network Change Risk Assessment Agent with description, metadata, actions, and documents

2. Catalogue view

The Agent Catalogue lists all agent cards. Each row surfaces key fields so you can browse and compare agents quickly.

Field Description
Name Agent display name
ID Unique identifier
Version Agent version
Project Project the agent belongs to
Category Functional grouping (for example NetOps, AIOps)
Description Summary of what the agent does
State Enabled or Disabled
Status Deployment status

3. Creating an agent card

You create a catalogue entry by packaging an existing agent into a card. You can add new cards and edit cards you already have.

Configuration

When creating or editing an agent card, you define:

Field Description
Agent Name Name shown in the catalogue
Description What the agent does
Version Card or agent version
Category How the agent is grouped
Project Owning or associated project

4. Upload agent (ZIP)

The Agent Catalogue supports uploading agents as packaged ZIP files so you can import and reuse agents across projects.

Package contents

A ZIP can include:

  • Agent metadata
  • Workflows
  • Personas
  • Supporting assets — Documents, images, and other files

Teams use packages to share and reuse prebuilt agents.

Configuration

When you upload an agent, you provide:

Field Description
AI Project Name Target project where the agent is imported
Agent ZIP File Packaged agent file to upload

Behavior

  • The uploaded package is validated and processed.
  • Agent components are registered in the selected project.
  • A new Agent Card is created in the catalogue.
  • Associated workflows, personas, and assets become available for that project.

Usage

  • Import agents from another environment
  • Share agents across teams
  • Deploy prebuilt agent packages

5. Download agents (ZIP)

The Agent Catalogue supports exporting agents as packaged ZIP files for reuse, sharing, or backup.

Overview

You can select one or more agents from the catalogue and download them as a ZIP package.

The export contains:

  • Agent metadata
  • Associated workflows
  • Personas
  • Supporting assets — Documents, images, videos, and other files

Selecting agents

In the download flow:

  • A list of available agents is shown.
  • You can select one agent or multiple agents.
  • Only selected agents are included in the export.

Behavior

  • Selected agents are packaged into a single ZIP file.
  • All associated components and assets are bundled together.
  • The ZIP can be imported later using Upload agent (ZIP).

Usage

  • Share agents across teams or environments
  • Back up agent configurations
  • Migrate agents between projects

6. Associated components

An agent card can link the following building blocks:

Persona

Attach one or more personas that define how the agent behaves in conversation and execution.

Workflow

Link workflows that implement the agent’s logic and automation paths.

7. Supporting assets

You can attach resources that help users adopt the agent:

  • Documents — For example PDFs and guides
  • Images — Screenshots and diagrams
  • Videos — Walkthroughs and demos
  • Icons — Visual identity for the card

These assets clarify purpose, setup, and day-to-day usage.

8. Agent card details

Opening an agent card shows:

  • Description and metadata
  • Version and publisher details
  • Category and status
  • Linked workflows and personas
  • Attached documentation and media

9. Agent card actions

Each agent in the catalogue exposes actions to manage, operate, and run the agent.

Interact, workflows, and observe

From the agent card, you can:

  • Interact with the agent — Opens Fabaio (the conversational copilot) using the persona and LLM configured for that catalogue entry. You get a direct chat with the agent: ask questions, refine tasks, and steer the interaction in natural language while the agent responds with the tools and behavior tied to that persona.
  • View agentic workflows — Open workflows tied to the agent
  • Observe — Open execution insight and observability views for that agent

Note

Interact with the agent launches Fabaio. For the full copilot experience—interface, MCP tools, cached documents, model and persona controls, and related features—see Fabaio – Conversational AI Copilot.

Edit

Opens the agent card for editing. You can update metadata (name, description, version, category), add or remove linked personas and workflows, and manage supporting assets such as documents, images, and videos.

Download

Export this agent as a ZIP package. The export includes agent metadata, workflows, personas, and assets. You can reuse the ZIP or import it into another project (see Upload agent (ZIP)). For exporting multiple agents from the catalogue list in one ZIP, see Download agents (ZIP).

View conflicts

Opens the Conflicts view for this agent so you can inspect and fix dependency or configuration issues. Full behavior, Pending status, and resolution steps are described in Conflict resolution.

Deploy agent

Deploys the agent so it is available for use: registers it for execution and activates associated workflows and personas according to your environment.

Disable

Turns the agent off without removing it. Usage of the agent is blocked while configuration and assets stay in place for later re-enable.

Delete agent

Removes the agent from the catalogue permanently, including the agent card and associated configuration.

Warning

Delete cannot be undone. Use Disable if you only need to stop usage temporarily.

10. Conflict resolution

When you import or upload agents, the system validates dependencies and configuration. If something is missing, mismatched, or incompatible, the agent is marked Pending until the issues are cleared.

What causes conflicts

Conflicts can occur when:

  • Referenced personas are not available in the target project
  • Required workflows are missing or do not match what the package expects
  • Dependencies in the agent package do not align with the current environment
  • Configuration from the upload is incomplete or incompatible with the project

Conflict status

  • Agents with unresolved issues appear as Pending
  • Pending agents cannot be deployed or executed until conflicts are resolved

Resolving conflicts

The platform provides a Conflicts view for each agent. From that view you can:

  • See which components are missing or mismatched
  • Map unavailable personas or workflows to ones that exist in the project
  • Adjust configuration so it matches the current project
  • Validate dependencies before you deploy

You apply changes directly; updates take effect in real time as you resolve items.

After resolution

When all conflicts are cleared:

  • The agent is ready for deployment
  • Status moves from Pending to Deployable
  • You can deploy the agent and use it normally

Outcomes

Conflict resolution helps ensure imported agents match the environment:

  • Missing dependencies are surfaced automatically
  • You can correct and map components in real time
  • Invalid or incomplete deployments are avoided before you go live
  • Agents reach a consistent, execution-ready state

11. Summary

The Agent Catalogue delivers:

  • Standardized packaging — One card per agent with clear structure
  • Discovery and reuse — Find and share agents across teams
  • Rich context — Documentation and media alongside technical components
  • Operational continuity — Interaction paths and observability from the same entry point

Upcoming: A2A client catalogue

Planned capability

In upcoming releases, Agent Catalogue is planned to act as an A2A client layer: each catalogue entry can be exposed as an A2A agent so that other A2A agents can invoke it or exchange messages with it. This extends the Agent Card and A2A discovery model to cross-agent collaboration and orchestration.