Skip to content

Agentic Automation Canvas (AAC)

What it is

The Agentic Automation Canvas (AAC) is a structured framework and open-source tool for the prospective design, governance, and evaluation of agentic AI systems. It provides a machine-readable "project contract" that bridges the gap between high-level user expectations and technical implementation.

What problem it solves

It addresses the Expectation-Realisation Gap: the systemic discrepancy where users expect high productivity gains from AI (e.g., 24% speedup) but often experience a decrease (e.g., 19% slowdown) due to unmeasured verification burdens, workflow friction, and human oversight costs. AAC requires explicit quantification of these factors during the planning phase.

Where it fits in the stack

Layer 6: Agents & Orchestration — specifically as a Planning & Design Framework that precedes and guides implementation.

Typical use cases

  • Full Project Planning: Designing end-to-end agentic workflows, including complex components like deterministic schedulers and LLM routers.
  • Governance & Compliance: Documenting data access, sensitivity, and staging for institutional or clinical AI deployments.
  • AI Coding Integration: Generating implementation-ready instructions for AI coding assistants.

Strengths

  • Six-Dimensional Framework: Covers Scope, User Expectations, Feasibility, Governance, Data Access, and Outcomes.
  • RO-Crate Export: Generates FAIR-compliant, machine-interoperable metadata packages following W3C and Schema.org standards.
  • AI-Ready Output: Automatically generates an AGENTS.md file that translates the project specification into structured instructions for tools like Cursor or GitHub Copilot.
  • Privacy-First: Fully client-side web application with real-time validation; data never leaves the browser.

Limitations

  • Beta Version: Currently in version 0.14.0; the schema and documentation are subject to change before 1.0.0.
  • Human Input Required: The quality of the output depends on the accuracy of the user's benefit quantification and feasibility assessments.

When to use it

  • Before starting the development of a new agentic system to ensure alignment between users and developers.
  • When you need to justify the ROI of an AI automation project by factoring in human-in-the-loop costs.
  • For complex projects requiring cross-backend model coordination and structured planning.

When not to use it

  • For trivial, single-prompt AI tasks that do not require tool-calling or multi-step reasoning.
  • When a project's goals and constraints are already fully documented in a compatible machine-readable format.

CLI examples

# AAC is a web-based Vue.js tool, no direct CLI for canvas creation.
# However, the exported RO-Crate can be validated via:
rocrate validate ./exported-crate/

# To self-host the AAC application locally:
git clone https://github.com/slolab/agentic-automation-canvas.git
npm install && npm run dev

Getting started

Online Access

Access the live version of the canvas at aac.slolab.ai. No installation is required as it is a client-side web application.

Basic Workflow

  1. Define Scope: Enter the primary goals and constraints of your agentic project.
  2. Quantify Benefits: Input expected productivity gains and human verification costs.
  3. Export Metadata: Save your project as an AGENTS.md file or an RO-Crate package for interoperability.

API examples

{
  "@context": "https://w3id.org/ro/crate/1.1/context",
  "@graph": [
    {
      "@id": "./",
      "@type": "Dataset",
      "name": "My Agentic Project",
      "description": "An automated workflow for document synthesis",
      "hasPart": [
        { "@id": "AGENTS.md" }
      ]
    },
    {
      "@id": "AGENTS.md",
      "@type": "File",
      "about": "Project implementation contract"
    }
  ]
}

Licensing and cost

  • Open Source: Yes (Apache License 2.0)
  • Cost: Free
  • Self-hostable: Yes (Vue.js application)

Sources / References

Contribution Metadata

  • Last reviewed: 2026-05-20
  • Confidence: high