MadeForMachine serviceEarly access

MFM Spec

A specification harness that pushes back.

MFM Spec turns specification from a document into an active organizational capability. It helps people and agents create, challenge, improve, and validate consistent specifications against an explicit schema, project rules, and active criteria. Start with rough intent or import an existing specification, then keep improving it without losing the reasoning that made it coherent.

MadeForMachine service

How MFM Spec fits

Your agent

Supplies the judgment: it interrogates intent, challenges decisions, and proposes coherent changes.

The service

Keeps the canonical spec, exposes precise reads and mutations, validates the graph, and records its history.

Your organization

Owns the intent, criteria, project rules, and portable specification artifact.

How it plugs in

Install the MFM Spec skill, then use authenticated, project-scoped tools through the shared MCP endpoint.

$ examples

Inspect four canonical, read-only specs before creating an account. Each example shows what MFM Spec can express and states what it does not prove.

open the gallery →

$ what changes

Most specifications describe what someone once intended. MFM Spec participates in what happens next. Your agent reads the same responsibilities, behavior, criteria, questions, and review evidence whenever the design changes.

That lets the agent challenge assumptions, expose contradictions, trace consequences, and refuse structurally invalid changes before they become implementation problems. The specification becomes an active organizational capability, not another document waiting to become stale.

$ authority

The agent may challenge the user, but it does so on the organization's authority: using the intent, criteria, rules, and evidence the organization chose. MFM Spec does not run a model or take over the build. Your agent supplies judgment; the harness supplies memory, precise operations, and deterministic boundaries.

$ graph

Four node types keep the model small enough to understand and rich enough to challenge. Decisions and reasoning live inside those nodes instead of becoming disconnected tickets.

components

Where does responsibility live?

features

What can an actor do or observe?

criteria

What must the design preserve, pursue, or avoid?

evaluations

What did we learn by reviewing something?

$ criteria in motion

A criterion is design pressure, not a ceremonial checkbox. Applying a catalog template copies it into the project; from then on the project owns it and the agent actively uses it while proposing and reviewing changes.

Catalog template

Reusable starting point

Project criterion

Owned and scoped locally

Active guidance

Applied across the design

$ two places to keep it

localskill

Your agent writes one manifest plus Markdown node files in the project, then validates the graph with the bundled linter.

hostedMCP

A database-backed canonical spec with revision history, deterministic mutations, minimal-context reads, and portable export.

$ shape

my-system/
  mfm-spec.yaml
  system.md
  components/
    api.md
    identity.md
  features/
    sign-in.md
  evaluations/
    sign-in-v1.md

Local files make the artifact tangible. Stable ids and typed relationships make the same artifact economical for an agent to query and validate.

$ integrity

1

Exactly one root component, named by the manifest.

2

Every parent, dependency, feature touch, and evaluation subject resolves.

3

Component and dependency graphs stay acyclic.

4

Responsibilities, intents, and summaries stay single-sentence and machine-sliceable.

5

Reasoning lives in the body, not only in frontmatter labels.

$ current boundary

MFM Spec creates, stores, evolves, validates, and explains specifications. It does not automatically inventory an organization, prove that the specification describes reality, verify code against the graph, or author through a browser chat.

$ distinct MFM service

Atlas is not another Spec mode. It is a separate product-intelligence service that helps agents compare external providers using grounded capability evidence.

explore Atlascompare the SDD landscape