Manifesto

Appearance

Shared Domain Model ​

Define the domain once, then let UI, routes, scripts, and agents call the same actions and read the same Snapshot shape.

The Core Idea ​

The important idea is not "multi-actor support" or "AI integration" as separate features. The important idea is that there is one domain model, and every consumer works against it:

  • the same MEL domain declaration
  • the same action names and input shapes
  • the same snapshot() read shape
  • the same deterministic transition rules

This is what makes Manifesto useful underneath app state, server routes, and agent tools without becoming a state management library or an AI framework.

One Domain, One Starting Runtime ​

text
Base runtime -> activate() -> action.x.submit(...) -> next Snapshot

The base runtime is the place to start. It is enough for UI, backend routes, scripts, and trusted agent tools.

The surfaces can be:

  • a React UI
  • a REST/GraphQL API handler
  • a CLI script
  • a background job
  • an LLM-powered agent

The contract stays the same. In ordinary app code, every surface reads through snapshot() and writes through app.action.<name>.submit(...).

When the product later needs history, restore, audit, approval, or policy, you can add Lineage or Governance around the same domain. That is a later decision, not part of the first app path.

text
base app -> optional Lineage/Governance extensions -> Snapshot plus extension records

A Small Example ​

typescript
import { createManifesto } from "@manifesto-ai/sdk";
import TodoMel from "./domain/todo.mel";

const app = createManifesto(TodoMel, {}).activate();

// UI surface
await app.action.addTodo.submit("Human-created todo");

// Agent surface - same action contract
await app.action.addTodo.submit("Agent-created todo");

Optional extension surfaces use the same domain actions, then add history, restore, audit, approval, or policy behavior around the base runtime.

Why This Matters ​

Without a shared domain model, teams end up with:

  • one state model for UI
  • another for the backend
  • another for automation
  • another for agents
  • another for audits

Each representation is a new inconsistency opportunity. Manifesto collapses those into one action and Snapshot contract.

Optional Approval And History ​

Sometimes the shared domain model is enough. Sometimes you need stronger controls:

  • explicit actor identity
  • approval rules
  • proposal review
  • audit history

Start with When You Need Approval or History. Read Lineage Records only when you want the Lineage record model behind history, restore, audit, or approval/history runtimes.

See Also ​