# indexio [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://github.com/IOServicesLabs/indexio  
**GitHub Stars:** 8  
**npm Downloads (last month):** 1267  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/indexio

## Description
Code search, symbols and call graph for AI coding agents. Answers in milliseconds.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `uvx` (confidence: high):

```json
"mcpServers": {
  "indexio": {
    "command": "uvx",
    "args": ["indexio-cli"],
    "env": {
      "INDEXIO_DATA_DIR": ""
    }
  }
}
```

**Requires environment variables:** `INDEXIO_DATA_DIR` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What indexio MCP server does

The indexio MCP server turns indexed source code into tools for AI coding agents. It can search for text, identifiers, regular-expression matches, or natural-language descriptions, then expose the results through MCP. Search supports lexical methods, semantic vector search over syntax-aware code chunks, and hybrid ranking that combines lexical and semantic results.

The index also records symbol definitions and call relationships. This lets an agent locate a symbol, identify callers, inspect a file outline, read a selected span, and estimate the effect of changing a symbol or a diff. The indexed data remains on the machine in the configured data directory.

## How it works

Run `indexio add` against a folder to register every Git repository below it, including repositories at nested levels. A non-Git folder can also be indexed as-is. The command builds the index immediately. By default, data is stored in `~/.indexio`; use `--data-dir` or `INDEXIO_DATA_DIR` to select another location.

After indexing, an MCP client starts the server with `indexio mcp`. The client sends tool requests to the local process, and responses come from the on-disk index. The same indexed data can be copied to another machine. `indexio sync` can be run from a cron job or timer to keep the index current.

## Setup and configuration

Install the binary with the Linux or macOS installation script, the Windows PowerShell script, npm, pip, Docker, or Cargo. The pip package is named `indexio-cli`, while the installed command is `indexio`. Git must be available on `PATH`.

A basic setup is:

1. Run `indexio add ~/code`, replacing the path with the folder containing repositories.
2. Register `indexio mcp` in the MCP client's configuration, using an absolute binary path if necessary.
3. Restart the agent session so it loads the new server.
4. Check the installation with `indexio stats`, `indexio search`, and `indexio usage`.

Claude Code can instead use `indexio setup claude`, followed by `indexio hook install`. The optional hooks route file reads and grep operations through the index.

## Tools and capabilities

The MCP tool set includes:

- `code_search` for indexed code search
- `code_grep` for grep-style matching
- `find_symbol` for symbol lookup
- `who_calls` for caller discovery
- `file_outline` for file structure
- `read_span` for reading a selected file range
- `impact_of_symbol` and `impact_of_diff` for change analysis
- `recall` for retrieving relevant indexed context

The command-line interface also provides search, repository statistics, synchronization, and usage reporting. The HTTP API offers another way to use the index outside MCP.

## Limitations and notes

The indexio MCP server does not provide a hosted database or external indexing service; its index is stored as local files. The README does not specify a client-independent authentication model for the HTTP API. Homebrew and winget packages are not currently available, so installation uses a script or another listed package method. Existing agent sessions must be restarted after setup or upgrade to use the new server version.

_Full upstream README: https://allmcps.com/mcp/indexio/readme_