@kr-pro-labs/clonevibe-mcp

Surgical AST Feature Slicer, Zero-Dependency Chameleon Code Grafter & 1:1 Golden Parity MCP Server for Vibe Coding.

During Vibe Coding, engineers frequently discover a specific algorithm, table parser, retry scheduler, physics routine, or state machine buried inside a massive external reference (a 50,000-line GitHub repository, an npm:/pypi:/cargo: package, or a minified JS bundle). Cloning the entire repository or installing bloated dependency trees pollutes your project.

clonevibe-mcp extracts only the minimal transitive AST closure of the target feature, replaces third-party dependencies with lightweight zero-dependency inline micro-shims, adapts the code to your host project's language and style conventions, and mechanically proves 1:1 behavioral equivalence via Golden I/O vectors.

Architecture Pipeline

flowchart LR
    A["Reference Input\n(GitHub / Local / npm: / pypi: / cargo: / JS Bundle)"] --> B["1. ingest_reference\nMultiSourceIngestor"]
    B --> C["2. locate_feature_entrypoints\nSymbol & Centrality Ranker"]
    C --> D["3. slice_engine_graph\nTransitive AST Closure Slicer"]
    E["4. profile_target_vibe\nChameleon Style Scanner"] --> F["5. synthesize_chameleon_graft\nZero-Dep Inliner & Code Synthesizer"]
    D --> F
    F --> G["6. verify_engine_parity\n1:1 Golden I/O Parity Verifier"]

Key Capabilities

All-in-One Multi-Source Ingestion: Ingests local codebases, GitHub URLs, package registry specs (npm:, pypi:, cargo:), inline file maps, and single-line minified JS bundles (with automatic token-aware structural formatting).
Polyglot Transitive AST Closure Slicing: Extracts top-level functions, classes, structs, interfaces, and call edges across TypeScript, JavaScript, Python, Rust, and Go, pruning 80%–98% of unrelated repository boilerplate.
Zero-Dependency Micro-Shim Catalog: Automatically replaces external imports (lodash, uuid, axios, node-fetch, chalk, etc.) with self-contained standard-library inline primitives (externalDepCount === 0).
Chameleon Style & Cross-Language Assimilation: Profiles your workspace (camelCase vs snake_case vs PascalCase, esm vs commonjs, quotes, indentation, error handling) and synthesizes drop-in modules and Clean-Room Algorithmic Blueprints.
1:1 Golden Parity Verification: Executes empty, numeric boundary, Unicode/CJK, malformed, and deterministic seed Golden Vectors and emits a ready-to-run host unit test file (vitest, pytest, cargo test, or go test).

Quick Start & MCP Client Configuration

Install & Build from Source

git clone https://github.com/KR-PRO-Labs/clonevibe-mcp.git
cd clonevibe-mcp
npm install
npm run build
npm test

Claude Desktop / Cursor / Windsurf / Antigravity (`mcp_config.json`)

{
  "mcpServers": {
    "clonevibe-mcp": {
      "command": "npx",
      "args": ["-y", "@kr-pro-labs/clonevibe-mcp"]
    }
  }
}

MCP Tools Reference

Tool Name	Description
`ingest_reference`	Hydrates a GitHub URL, local directory, registry spec (`npm:`, `pypi:`, `cargo:`), inline file map, or minified JS bundle into a `ReferenceSnapshot`.
`locate_feature_entrypoints`	Scores and ranks symbols in a snapshot against natural-language feature queries and call centrality.
`slice_engine_graph`	Computes the minimal transitive AST call/type closure from entrypoint symbols and matches zero-dep shims.
`profile_target_vibe`	Inspects a target workspace or code sample to detect language, module system, casing, quotes, indentation, and error-handling idiom.
`synthesize_chameleon_graft`	Generates a zero-dependency adapted code module and Clean-Room Algorithmic Blueprint in the target language.
`verify_engine_parity`	Verifies 1:1 Golden I/O parity, confirms zero external dependencies, and outputs a runnable test suite.

MCP Resources

clonevibe://catalog/zero-dep-shims — JSON catalog of zero-dependency inline shims across TypeScript, JavaScript, Python, Rust, and Go.
clonevibe://guide/chameleon-workflow — Markdown guide for executing the 6-stage surgical grafting pipeline.

License

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support