Carve MCP
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Carve MCPCarve image transforms from REA evidence, compare behavior, save a reusable function folder."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
reverse-engineering-carve
Extract investigated software behavior into reusable local capabilities.
reverse-engineering-carve (RECarve) is an independent companion MCP for REA. Your agent uses REA to investigate software, then carve to define a feature, extract or rebuild it, verify behavior and save a reusable function folder. REA and carve run as separate MCP servers.
繁體中文 · Releases and detailed capabilities
Install
Requires Node22.19 in the22 line or Node24.11+. In the cloned repository:
npm ci
npm run build
node dist/cli.js --versionCLI: carve. MCP entry: carve. Project/package: reverse-engineering-carve. RECarve is a prose abbreviation. Release packages are available on GitHub; the project is not published to npm.
Related MCP server: spine
Add to an existing REA setup
Keep your working REA installation and MCP entry. Merge the following entry into your client's existing mcpServers map, replacing the absolute paths:
{
"carve": {
"command": "node",
"args": ["/absolute/path/to/reverse-engineering-carve/dist/cli.js", "serve", "--root", "/absolute/path/to/workspace"]
}
}Reload MCP connections. Your agent should see REA and carve separately. No REA reinstallation, additional LLM API key or website is required. Setup details.
Use
Use REA to investigate this application and carve to suggest reusable features. Show the evidence and unknowns. Once I select a feature, define its boundary, extract or rebuild it, compare concrete behavior and save a reusable local folder.
Describe where the feature is triggered, its inputs and expected outputs. The agent still needs to investigate dependencies; native/mobile support remains experimental, and a successful build does not prove original behavior equivalence. Execution-enabled actions require trusted code and authorization.
Release notes, capabilities and verification scope · MIT license
This server cannot be deployed
Maintenance
Related MCP Connectors
Repository evidence for agents before they adopt dependencies, enter codebases, compare, or merge.
Machine-native research commons for agent evidence, discovery, rooms, and bounded research quests.
- SpecitaOAuthcom.specita
Read-only release evidence for coding agents: findings, deliveries, source and runtime data.
Evidence infrastructure for agents: source-backed company verification and beta import assessment.
Related MCP Servers
- AlicenseAqualityBmaintenanceEnables coding agents to analyze a repository evidence-first, answering questions as verifiable claims with source evidence, targeted test observations, and reference resolution.53Apache 2.0
- AlicenseNot gradedqualityAmaintenanceProvides local-first repository architecture analysis with file-level proof, enabling agents to map codebases, locate implementations, trace call paths, and assess change impact with deterministic evidence.4 npm7MIT
- FlicenseNot gradedqualityAmaintenanceEnables coding agents to coordinate profilers, benchmark tools, trace processors, and direct local targets so that explicit artifacts or commands become bounded runtime evidence with recorded provenance, coverage, and limitations. Its analysis and capture tools require no workspace setup, and durable evidence preservation remains optional.-
- AlicenseNot gradedqualityBmaintenanceEnables coding agents to extract a source feature’s static dependency closure and adapt it into a target Git repository’s own data model and design, then carry later upstream fixes into the adapted copy with pinned provenance and frozen verification checks.MIT