src-to-kb

Turn any source code repo into a searchable, AI-queryable knowledge base inside Claude or Cursor.

Unverified

stdio (local)

No auth

JavaScript

View repo 34 npm

Add to your client

Copy the config for your MCP client and paste it into its config file.

Install / run

npm install -g @vezlo/src-to-kb

Paste into ~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "src-to-kb": {
      "command": "node",
      "args": [
        "/Users/admin/Documents/AI-Playground/src-to-kb/mcp-server.mjs"
      ],
      "env": {
        "OPENAI_API_KEY": "your-api-key-here"
      }
    }
  }
}

Step-by-step guides: Add to Claude Desktop · Add to Cursor · Add to Windsurf

Before you start

Node.js 14+ and npm
Claude Code, Claude Desktop, or Cursor with MCP support
The @vezlo/src-to-kb npm package installed (globally recommended)
OpenAI API key (OPENAI_API_KEY) — optional, only needed for embeddings and AI-powered answers

About src-to-kb

src-to-kb indexes a source code repository into a knowledge base (documents, chunks, and optional embeddings) and exposes that knowledge base to AI assistants over MCP. The MCP server is started via the bundled mcp-server.mjs (or the src-to-kb-mcp command) and gives Claude/Cursor tools to generate a KB, run natural-language search with AI-generated answers, get codebase statistics, list/filter indexed files, and find similar files. The same package also ships a CLI (src-to-kb, src-to-kb-search, src-to-kb-upload) and a REST API server (src-to-kb-api) with Swagger docs, plus Notion import and optional integration with an external assistant-server for production vector search.

Tools & capabilities (7)

generate_kb

Generate a searchable knowledge base from a source code repository by chunking, cleaning, and optionally creating OpenAI embeddings. Params: repoPath (required), outputPath, chunkSize, chunkOverlap, excludePaths, includeExtensions, generateEmbeddings.

search_kb

Run an intelligent natural-language search across the knowledge base with AI-generated answers. Params: query (required), kbPath, limit, includeContext.

get_kb_stats

Return comprehensive statistics about an indexed knowledge base, including document/chunk counts, total size, language distribution, and file types. Param: kbPath.

list_kb_files

List all files indexed in the knowledge base, with optional filtering by programming language or file type (code/text/config/web). Params: kbPath, filterByLanguage, filterByType.

find_similar

Find files similar to a given file based on structure, language, and content patterns. Params: filePath (required), kbPath, limit.

extract_code

Extract specific code patterns from the codebase (e.g. all React component definitions).

update_kb

Update the knowledge base with new changes from the source code.

When to use it

Index an entire codebase so Claude/Cursor can answer questions about how it works
Onboard new developers by letting them ask natural-language questions about an unfamiliar repo
Investigate bugs or audit security by searching for relevant code across a large repository
Discover related files and similar implementations when refactoring
Prepare code as a knowledge base for RAG systems

Security notes

All processing happens locally on your machine; no code is sent to external servers except optional OpenAI calls when embeddings are enabled (set OPENAI_API_KEY). Knowledge bases are stored in your local filesystem. The software is dual-licensed (AGPL-3.0 for non-commercial use; a commercial license is required for commercial use).

src-to-kb FAQ

Do I need an OpenAI API key?

No. The server and basic keyword search work without one. An OPENAI_API_KEY is only required if you want to generate embeddings for semantic search or AI-generated answers; set it in the MCP server's env.

Is my code sent to external servers?

No. All indexing and search happen locally on your machine. The only outbound calls are optional OpenAI requests when embeddings are enabled. (Pushing to an external assistant-server is a separate, opt-in CLI feature.)

Which editors does it work with?

Any MCP-compatible client. The docs cover Claude Code (via `claude mcp add`), Claude Desktop (via an mcpServers config block), and Cursor.

How do I install and register the server?

Install the package with `npm install -g @vezlo/src-to-kb`, then either run `claude mcp add src-to-kb -- node <path>/mcp-server.mjs`, or add an mcpServers block pointing `node` at mcp-server.mjs to your client config and restart it.

What languages are supported?

JavaScript, TypeScript, Python, Java, C/C++, C#, Go, Rust, Ruby, PHP, Swift, Kotlin, Scala, and more, plus config/markup formats like JSON, YAML, XML, Markdown, HTML/CSS, and SQL.

#code-search #knowledge-base #rag #embeddings #code-indexing #code-analysis #openai #vector-search

Alternatives to src-to-kb

Compare all alternatives →

Playwright MCP Server

Developer Tools

24k

Microsoft's official browser-automation MCP using Playwright's accessibility tree (no vision model).

Featured

Verified

stdio (local)

No auth

TypeScript

12 tools

Updated 21 days agoRepo

Context7 MCP Server

Developer Tools

28k

Up-to-date, version-specific library documentation injected into your coding agent.

Verified

stdio (local)

API key

TypeScript

2 tools

Updated 25 days agoRepo

codebase-memory-mcp

Developer Tools

16k

Code intelligence engine that indexes repos into a persistent knowledge graph for AI coding agents.

Unverified

stdio (local)

No auth

14 tools

Updated 11 hours agoRepo

Compare src-to-kb with:

vs Playwright MCP Server vs Context7 MCP Server vs codebase-memory-mcp vs Serena