Give your AI coding agentsa map of your code
Nuanced is an MCP server that provides agents like Cursor and Claude Code with precise TypeScript call graphs for more accurate edits, faster builds, and lower token spend.
How to get started with Nuanced MCP (2 minutes)
Once you subscribe, we'll email your license key and installation instructions. If it's not in your inbox, check your spam folder. You'll receive the email within minutes. Keep your license key private.
Step 1: Configure npm registry
npm config set @nuanced-dev:registry 'https://nuanced-dev.nodejs.pub'
npm config set '//nuanced-dev.nodejs.pub/:_authToken' "YOUR-SUPER-SECRET-LICENSE"Step 2: Install Nuanced MCP
npm install -g @nuanced-dev/nuanced-mcp-tsStep 3: Configure MCP clients
For Cursor:
{
"mcpServers": {
"nuanced": {
"command": "nuanced-mcp-ts"
}
}
}For Claude:
claude mcp add nuanced nuanced-mcp-tsStep 4: Prompt your agent with Nuanced
Open Claude Code or Cursor and write your prompt as usual. Append "use Nuanced" to fetch call-graph and symbol context via MCP and pass it to your agent. Under the hood, Nuanced will run nuanced init to re-index the project graph and nuanced enrich <file> <function> to extract the relevant subgraph before responding to your request.
How it works
Agent-ready
Scan repo → build call graph
Select the slice your prompt needs
Pass to any LLM for accurate answers and code that compiles
Proven impact
Local-first, runs anywhere
Works with any LLM or coding workflow: OpenAI, Claude, Cursor, VS Code, even your CI pipeline.
Analysis never leaves your machine.
Built by developers who've been there
We spent years building code intelligence and static analysis at GitHub. Nuanced brings that rigor to AI coding agents: precise call graphs, measurable accuracy, and a local-first design. We use it on our own code every day and ship improvements weekly.