Pharaoh.so installs statistics

Pharaoh maps your codebase into a knowledge graph and gives AI coding tools full architectural context through MCP - every function, dependency, call chain, and blast radius in ~2K tokens instead of 40K.

The problem

AI coding tools read your codebase one file at a time. They burn 40,000+ tokens just exploring structure, miss cross-module dependencies, create duplicate functions that already exist, and write PRDs that reference code that isn't there.

Every session starts blind.

What Pharaoh does

Pharaoh parses your repos into a queryable knowledge graph. Your AI gets full architectural awareness before it touches anything.

16 MCP tools your AI can call directly:

• get_codebase_map - Full architecture overview in one query • search_functions - Find existing code before writing new code • get_blast_radius - See what breaks before you refactor • check_reachability - Verify functions are actually wired to endpoints • get_unused_code - Find dead code nobody calls • query_dependencies - Trace how modules connect • get_module_context - Everything about a module before modifying it • get_test_coverage - See what's tested and what isn't

How it works

• Install the GitHub App and select your repos • Pharaoh maps them into a knowledge graph (read-only - we never write to your code) • Add the MCP endpoint to Claude, Cursor, or any MCP-compatible tool • Your AI queries the graph instead of reading files blind

Graphs refresh automatically on every push to your default branch.

Languages

TypeScript and Python.

Who it's for

Developers and teams using AI coding tools who are tired of their AI guessing at architecture it could just know.