CKB is a feature-rich code intelligence MCP server offering a comprehensive toolset (80+) covering navigation, impact analysis, architecture, security, ownership, and quality. Its compound operations and presets significantly reduce AI token consumption. The main strengths are static analysis and multi-language support, though quality varies (Go best, PHP experimental). It requires indexing, which can become stale, and does not handle dynamic behavior. Overall, it is highly useful for refactoring, review, and AI-assisted development, but be mindful of static analysis limitations and index maintenance.
CKB (Code Knowledge Backend) transforms your codebase into a queryable knowledge base. It provides 80+ MCP tools, CLI, and HTTP API for symbol navigation, impact analysis, architecture mapping, dead code detection, secret scanning, code ownership, and more. Leveraging SCIP indexes, LSP, Git, and optional LIP semantic embeddings, it offers precise code intelligence for AI assistants or as a backend for CI/CD and custom tooling. It features compound operations like explore, understand, and prepareChange that reduce AI tool calls by 60-70%, and presets that cut token usage by up to 77%. CKB has tiered language support, with Go, TypeScript, JavaScript, and Python being best.
Options: 1) npm (recommended): npm install -g @tastehub/ckb; 2) Homebrew (macOS/Linux): brew tap SimplyLiz/ckb && brew install ckb; 3) Build from source: git clone https://github.com/SimplyLiz/CodeMCP.git && cd CodeMCP && go build -o ckb ./cmd/ckb. After installation, run ckb init in your project directory, optionally ckb index to generate SCIP index, then ckb setup to connect Claude Code. You can also manually add npx @tastehub/ckb mcp to your MCP config.
{
"mcpServers": {
"ckb": {
"command": "npx",
"args": [
"@tastehub/ckb",
"mcp"
]
}
}
}