JSPM

  • ESM via JSPM
  • ES Module Entrypoint
  • Export Map
  • Keywords
  • License
  • Repository URL
  • TypeScript Types
  • README
  • Created
  • Published
  • Downloads 79
  • Score
    100M100P100Q91333F
  • License MIT

ORBITAL — AI-Native Development Framework. Structural governance for AI coding agents.

Package Exports

    This package does not declare an exports field, so the exports above have been automatically detected and optimized by JSPM instead. If any package subpath is missing, it is recommended to post an issue to the original package (thinkorbital) to support the "exports" field. If that is not possible, create a JSPM override to customize the exports field for this package.

    Readme


    ◯ R B I T A L
    Structural governance for AI coding agents

    npm version license docs

    ORBITAL is an AI-native development framework that gives your coding agent structure, boundaries, and accountability. Instead of letting AI loose on your codebase, ORBITAL enforces a phased lifecycle where every change is scoped, proposed, approved, and verified.

    npm install -g thinkorbital
    cd your-project
    orbital liftoff

    Why ORBITAL

    AI coding agents are powerful but ungoverned. They hallucinate file paths, overwrite working code, ignore project conventions, and operate without audit trails. ORBITAL fixes this:

    • Scoped execution — AI only touches files it's been granted access to
    • Human gates — proposals require approval before any code is written
    • Isolated branches — every task runs in its own git worktree
    • Full audit trail — every decision, every file access, every phase transition is logged

    How It Works

    Every unit of work follows the ICPAEV lifecycle:

    Intent → Context → Proposal → Authorization → Execution → Verification
    Phase What happens
    Intent Define what you want built — objective, scope, success criteria
    Context Resolve which files the agent can read and write (CBAC envelope)
    Proposal AI generates a technical plan — approach, files, risk assessment
    Authorization Human reviews and approves (or rejects) the proposal
    Execution Agent writes code in an isolated worktree with enforced boundaries
    Verification Tests pass, criteria met, changes merge back to your branch

    Key Concepts

    Context-Based Access Control (CBAC)

    Every intent gets a context envelope — a precise manifest of which files the agent can read, which it can modify, and which are off-limits. This isn't advisory; it's enforced at the MCP gateway level.

    ┌─────────────────────────────┐
    │ Context Envelope             │
    │                              │
    │  ✓ writable: src/auth/*.ts   │
    │  ○ readable: src/types/*.ts  │
    │  ✗ denied:   .env, secrets/  │
    └─────────────────────────────┘

    Tiered Governance

    Tier Use case Governance
    T1 Quick fixes, small changes Auto-advance, minimal gates
    T2 Features, refactors Human authorization required
    T3 Architecture, design work AI-led design conversation → decomposed into sub-intents

    MCP Server

    ORBITAL includes a Model Context Protocol server that your AI agent connects to automatically. The server exposes governance tools — the agent can't skip phases, exceed its file boundaries, or self-authorize. Every tool call is traced.

    CLI Commands

    orbital liftoff          Initialize ORBITAL in any project
    orbital orbit            Create a new trajectory + intent
    orbital context          Resolve context envelope
    orbital propose          Generate AI proposal
    orbital authorize        Approve proposal for execution
    orbital execute          Begin isolated code execution
    orbital verify           Run verification checks
    orbital status           View all trajectories and intents
    orbital monitor          Live terminal dashboard
    orbital design           T3 design conversation
    orbital systems          Health check (--fix to auto-repair)

    Full command reference: thinkorbital.ai/docs/cli-lifecycle

    What liftoff Creates

    your-project/
    ├── .orbital/
    │   ├── config.yaml              # Project detection + settings
    │   ├── ORBITAL.md               # Agent instructions (auto-generated)
    │   ├── trajectories/            # Work streams
    │   ├── orbits/                  # Intent state + evidence
    │   ├── policy/                  # Security policies + capability profiles
    │   ├── context/                 # Layers, manifest, generated context
    │   ├── envelopes/               # CBAC access grants
    │   └── traces/                  # Gateway audit logs
    ├── .mcp.json                    # MCP server registration
    └── CLAUDE.md                    # Agent bootstrap instructions

    Works With

    • Claude Code (VS Code, CLI) — first-class MCP integration
    • Cursor — MCP server support
    • Any agent that speaks Model Context Protocol

    Requirements

    • Node.js 18+
    • Git 2.25+

    Documentation

    Full docs, architecture deep-dives, and guides at thinkorbital.ai

    License

    MIT