open format specification v1.0

.codeindex

An open format for structured code documentation.
Committed to git, shipped in packages, queried by AI agents.

.codeindex

├── index.json

├── files.jsonl

├── symbols.jsonl

└── texts.jsonl

Read the spec View on GitHub

AI agents waste tokens finding code

They grep, read files, grep again, backtrack. Thousands of tokens burned locating a function — or worse, they miss it and hallucinate.

Without .codeindex

grep -r "class Config" src/

cat src/main.py | head -50

grep -r "def __init__" src/main.py

cat src/main.py

... 4 tool calls, ~3000 tokens

With .codeindex

search("Config", scope: ["symbol"])

→ Config · class · src/main.py:22-45
→ Config.__init__ · method · :23-30
sig: def __init__(self, path: str, debug: bool = False)

... 1 tool call, ~200 tokens

Build and query with codeix

Index your codebase in seconds. Your AI agent finds the right function on the first try — no scanning, no guessing, no wasted tokens.

terminal

codeix build # parse source, write .codeindex

codeix serve # MCP server + file watcher

codeix -r ~/project build # build from a specific dir

MCP Tools

discovery

explore

Project structure: metadata, subprojects, and files grouped by directory.

Unified full-text search across symbols, files, and texts. FTS5, BM25-ranked, with scope/kind/path filters.

lookup

get_file_symbols

List all symbols in a file, ordered by line number.

lookup

get_children

Get direct children of a class, struct, or module.

graph

get_callers

Find all call sites and references to a symbol.

graph

get_callees

Find all symbols that a function calls or references.

Languages

Ship your index with your package

Include .codeindex in your package and every developer who depends on you gets instant navigation of your API — no setup, no re-indexing.

Author

Build the index

Run once on your source tree. Deterministic output, clean diffs.

codeix build

→

Publish

Ship with your code

Commit .codeindex to your repo or include it in your published package.

.codeindex

→

Consumer

Query instantly

AI agents navigate your API without parsing a single file.

search()

Git repos

npm

PyPI

crates.io

An open format anyone can read

Plain JSONL. Git-friendly diffs. No binary blobs, no proprietary databases. Any tool that can parse JSON can consume a .codeindex.

language-agnostictool-agnosticdeterministicgit-friendlyhuman-readable

symbols.jsonl

{"file":"src/main.py","name":"os","kind":"import","line":[1,1]}{"file":"src/main.py","name":"Config","kind":"class","line":[22,45]}{"file":"src/main.py","name":"Config.__init__","kind":"method","line":[23,30],"parent":"Config","sig":"def __init__(self, path: str, debug: bool = False)"}{"file":"src/main.py","name":"main","kind":"function","line":[48,60],"sig":"def main(args: list[str]) -> int"}

index.json

Manifest: version, project name, detected languages

files.jsonl

One line per source file: path, language, hash, line count

symbols.jsonl

Functions, classes, imports — with signatures and parent relationships

texts.jsonl

Comments, docstrings, string literals — searchable prose

Install

All channels install the same single binary. No runtime dependencies.

install

npx codeix # npm

uvx codeix # Python

cargo install codeix # Rust

brew install codeix # Homebrew

.mcp.json

{  "mcpServers": {    "codeindex": {      "command": "npx",      "args": ["codeix"]    }  }}