Four pieces, one pipeline
MCP server for NativeCode Audit
Lets Claude and other MCP clients run live SEO, AEO and GEO audits through app.nativecode.ai.
llms.txt and Markdown mirror
A curated llms.txt, the full site as one text file, and a clean Markdown copy of every page.
RAG over NativeCode content
Chunks of the Markdown mirror, MiniLM embeddings and a hybrid vector and BM25 index. Try it below.
Agentic DOM audit
How well browser agents and screen readers can perceive and operate each nativecode.ai template.
Search nativecode.ai with retrieval
Your question is embedded in this browser with the same MiniLM model that built the index, then matched against every chunk by meaning and by keywords. The first search downloads the model (about 23 MB).
Connect the MCP server
git clone https://github.com/nativecodeai/nativecode-geo
cd nativecode-geo/mcp-server && npm install && npm run build
claude mcp add nativecode-audit -- node "$PWD/dist/index.js"
Then ask: "Audit nativecode.ai for AI search readiness and give me the quick wins."