MDAAIRepository documentation

MDAAI

MDAAI is a repository-based operating protocol for human-directed AI development. It puts instructions, project scope, tasks, decisions and verification evidence in files so an agent can know what it may change, what is current and what supports a result.

By Eris Margeta Kurdali · Repository

A protocol kept with the project

The repository carries the working agreement between its operator and development agents. Instructions define constraints; the project plan defines scope; task records preserve current work and evidence. These are ordinary Markdown and JSON files, read and maintained alongside code—not a replacement for the application or its tests.

See the files →Follow one task →

Continuity without duplicate state

A new session should be able to find the authorized task, its acceptance criteria, current evidence and next action without reconstructing a conversation. In MDAAI 2.0, Project Elaboration owns scope and sequence; TASKS.json owns task state. Keeping each fact with one owner avoids conflicting status checkboxes, parallel registries and repeated work.

How the files work together →

What adoption adds—and does not add

The portable 2.0 core is a root AGENTS.md, five governance documents, project-owned Project Elaboration and a fresh TASKS.json. CLAUDE.md may point to AGENTS.md. Adoption is an explicitly authorized, project-by-project merge and migration, not a universal installer. The source package’s research code, dashboards, studies and existing task history are not copied into every adopted repository.

Exact 2.0 adoption boundary →

Two repository workflows

The first-generation template uses an elaborated project plan, Work Orders, a work-order registry, corrective records and architecture decisions. MDAAI 2.0 concentrates routine work in one task registry and adds separate records only when they carry necessary context. A migration reconciles contradictory instructions and preserves completed history; it does not simply add a second protocol on top.

First-generation MDAAI →MDAAI 2.0 →

Where Onion fits

In this repository, Onion Governance is the broader governance system; MDAAI 2.0 is its repository and coding-governance layer. That relationship is this project’s declared scope, not an extra file all MDAAI adopters must install. The Onion implementation and its laboratory evidence are distinct from the portable instructions documented here.

Instructions are not automatic enforcement

Markdown instructions guide agent behavior. Structural validators can check fields, references and consistency; they cannot establish that a claim is true, grant authorization or prove a product works. Actual tests and relevant runtime observation support acceptance. Automated revision/publication software is a separate implementation with separate approvals—not a consequence of reading AGENTS.md.

A technical overview of the original template and MDAAI 2.0. The endpoint example is illustrative, not a product implementation claim.

Have Trouble Reading?

Here's MDAAI in 91 seconds.

English narration and captions · 1080 × 1080 · 1 min 31.3 sec. Download captions (SRT).

Source references

Based on original contracts and templates. These are source locations, not installed-file claims. No private source archives are served.

  • MDAAI 2.0: AGENTS.md
  • MDAAI 2.0: CLAUDE.md
  • MDAAI 2.0: docs/ADOPTION.md
  • MDAAI 2.0: scripts/validate.py
  • MDAAI 2.0: PROJECT-INTERNAL/GOVERNANCE/AUTHORITY.md
  • MDAAI 2.0: PROJECT-INTERNAL/GOVERNANCE/ENGINEERING.md
  • MDAAI 2.0: PROJECT-INTERNAL/GOVERNANCE/EVIDENCE.md
  • MDAAI 2.0: PROJECT-INTERNAL/GOVERNANCE/REASONING.md
  • MDAAI 2.0: PROJECT-INTERNAL/GOVERNANCE/RECORDS.md
  • AGENTS.md
  • MDAAI 2.0: PROJECT-INTERNAL/MANAGEMENT/PROJECT-ELABORATION.md
  • MDAAI 2.0: PROJECT-INTERNAL/MANAGEMENT/TASKS.json

Search documentation

Search runs locally. No query leaves your browser.

↑ ↓ move · Enter open · Esc close