Skip to content

Agentic Automation Canvas (AAC)

Beta Release

Version 0.10.1 (Beta): This specification is currently in beta testing. The schema and documentation may change before the stable 1.0.0 release. Feedback and contributions are welcome!

The Agentic Automation Canvas (AAC) is a framework for designing and documenting agentic automation solutions driven by generative AI. Agentic systems are designed to replace human judgment in a reliable way, requiring a control inversion where the system takes command of tasks while humans step back accordingly. The AAC provides a structured approach to capture project metadata, user expectations, feasibility considerations, governance requirements, and outcomes—all while generating machine-readable outputs that integrate with FAIR Digital Object ecosystems.

The canvas serves simultaneously as a framework for implementation, a checklist for relevant aspects that need consideration, and a machine-readable specification for your agentic automation task. Once completed, the canvas can be exported as an interoperable RO-Crate that guides development through planning, prototype, and deployment stages.

Quickstart

  1. Write a canvas: Use the interactive web application to fill out your agentic automation canvas, capturing project details, requirements, feasibility, governance, and outcomes.

  2. Validate: The web form validates your data automatically as you fill it out. If you're working with canvas JSON files directly (outside the web form), use the validator to ensure compliance with the AAC specification.

  3. Download: Download your completed canvas as an RO-Crate package (ZIP file). The web form exports your data as a standards-compliant RO-Crate that can be shared, archived, and integrated with research infrastructure following FAIR principles.

Canonical Identifiers

The canonical base namespace for the Agentic Automation Canvas is:

https://w3id.org/aac/

All schema artifacts, examples, and documentation are served from this stable namespace:

  • Schema: https://w3id.org/aac/schema/aac.schema.json
  • Specification: https://w3id.org/aac/spec/
  • Examples: https://w3id.org/aac/examples/

Namespace Policy

Stability Promises:

  • Beta Phase (0.9.x): During beta testing, the schema structure may change. Breaking changes will be documented in the changelog.
  • Stable (1.0+): After 1.0.0, schema $id URIs, property names, required fields, and core structure will remain stable within major versions.
  • Versioned Changes: Breaking changes will only occur in major version increments (e.g., 1.0 → 2.0), with deprecation notices in advance.
  • Additive Evolution: Minor versions may add optional fields, new controlled vocabularies, or extended examples without breaking existing implementations.
  • Documentation: Specification documentation may be updated for clarity without changing the schema structure.

What May Change:

  • Documentation wording and examples (for clarity)
  • Additional optional fields in minor versions
  • Extended controlled vocabularies
  • New example files

What Won't Change:

  • Core property names and types
  • Required field definitions
  • Schema $id URIs (within major versions)
  • JSON Schema structure and validation rules

Resources

Specification & Documentation

  • Specification: Complete specification overview
  • Core Concepts: Key concepts (canvas, task, contract, evaluation, provenance)
  • Conformance: Conformance requirements, versioning, and deprecation policy

Schema & Artifacts

  • JSON Schema: Machine-readable schema definition (available after deployment)
  • YAML Schema: YAML representation of the schema (available after deployment)
  • Schema Reference: Auto-generated reference documentation

Examples

Tools & Validation

Community & Support

Standards Compliance

The Agentic Automation Canvas follows established standards:

  • RO-Crate 1.1: Research Object Crate specification
  • Schema.org: Project, ResearchProject, CreativeWork types
  • W3C DCAT: Data Catalog Vocabulary for datasets
  • W3C PROV-O: Provenance Ontology for activities
  • P-Plan: Plan Ontology for user expectations
  • FRAPO: Funding, Research Administration & Projects Ontology
  • DUO: Data Use Ontology for data restrictions

License

Licensed under the Apache License, Version 2.0. See the LICENSE file for details.