Skip to main content

High-performance local search and indexing engine for large codebases

Project description

Sari – Local Code Search/Indexing MCP Server

한국어 설명(Korean)

Sari is a local code search/indexing MCP server designed for fast, local indexing and search across large codebases. It defaults to MCP stdio integration, supports multi-workspace collection, and keeps configuration and data paths clearly separated.


1. Overview

Sari indexes large codebases locally and provides fast search capabilities via the Model Context Protocol (MCP). It ensures your source code stays on your machine while providing powerful search, symbol navigation, and call graph analysis to LLMs.


2. Installation

2.1 Recommended: uv tool (Global-like Utility)

Since Sari is an MCP server utility used across multiple projects, we strongly recommend using uv tool. It keeps the environment isolated while providing a stable, single binary path for all your workspaces.

  • Pros: Prevents redundant installations, stable command path for MCP config, and easy CLI access.
uv tool install sari

After installation, find the absolute path for your MCP settings:

which sari
# Example: /Users/yourname/.local/bin/sari

2.2 Optional: venv install (Per-project Isolation)

Use this only if you want to lock Sari to a specific project. Note that you'll need to install it in every workspace, and you must update your MCP command path whenever you switch projects.

uv venv .venv
source .venv/bin/activate
uv pip install sari

3. Runtime Modes

3.1 stdio (MCP Recommended)

The stdio transport runs as a daemon proxy for high performance.

Start the daemon (required for stdio):

sari daemon start -d

3.2 HTTP API

Start the HTTP API server for access via browser or other HTTP clients.

sari --transport http --http-api-port 47777

Health check:

curl http://127.0.0.1:47777/health

4. MCP Client Setup

Configuration depends on your installation method. Replace yourname with your actual system username.

4.1 Gemini CLI (~/.gemini/settings.json)

A. If installed via uv tool (Recommended)

{
  "mcpServers": {
    "sari": {
      "command": "/Users/yourname/.local/bin/sari",
      "args": ["--transport", "stdio"],
      "env": {
        "SARI_CONFIG": "/abs/path/to/workspace/.sari/config.json"
      }
    }
  }
}

B. If installed via venv

{
  "mcpServers": {
    "sari": {
      "command": "/abs/path/to/project/.venv/bin/python",
      "args": ["-m", "sari", "--transport", "stdio"],
      "env": {
        "SARI_CONFIG": "/abs/path/to/workspace/.sari/config.json"
      }
    }
  }
}

4.2 Codex CLI (~/.codex/config.toml)

A. If installed via uv tool (Recommended)

[mcp_servers.sari]
command = "/Users/yourname/.local/bin/sari"
args = ["--transport", "stdio"]

[mcp_servers.sari.env]
SARI_CONFIG = "/abs/path/to/workspace/.sari/config.json"

B. If installed via venv

[mcp_servers.sari]
command = "/abs/path/to/project/.venv/bin/python"
args = ["-m", "sari", "--transport", "stdio"]

[mcp_servers.sari.env]
SARI_CONFIG = "/abs/path/to/workspace/.sari/config.json"

5. Data & Install Paths

  • Global DB: ~/.local/share/sari/index.db
  • Global Registry: ~/.local/share/sari/server.json
  • Logs: ~/.local/share/sari/logs
  • Workspace Config: <workspace>/.sari/config.json or <workspace>/sari.json
  • Global Config: ~/.config/sari/config.json

6. Multi-workspace

6.1 CLI

sari roots add /path/to/workspaceA
sari roots add /path/to/workspaceB
sari roots list

6.2 Config file

{
  "workspace_roots": [
    "/path/to/workspaceA",
    "/path/to/workspaceB"
  ]
}

6.3 Collection Example

Register multiple workspaces in the global config and run.

# Register roots
sari roots add /Users/user/Repo1
sari roots add /Users/user/Repo2

# Run with global config
SARI_CONFIG=~/.config/sari/config.json sari --transport stdio

7. Indexing Policy

Sari determines indexing roots in this order:

  1. workspace_roots from global/workspace config
  2. SARI_WORKSPACE_ROOT environment variable
  3. rootUri/rootPath from MCP initialization
  4. Fallback to current working directory (CWD)

8. Configuration Reference

Key Description Default
workspace_roots Multi-workspace roots [CWD]
include_ext File extensions to index .py, .js, .ts, .java, ...
exclude_dirs Excluded directories .git, node_modules, .venv, ...
max_depth Max directory depth 20
scan_interval_seconds Auto-scan interval (sec) 180

9. Call Graph Options

call_graph becomes more stable when you provide explicit scope:

  • repo: force repository scope
  • depth: search depth (default: 2)
  • include_paths/exclude_paths: glob filters

10. Troubleshooting

Please see docs/TROUBLESHOOTING.md when issues occur.


11. Maintenance

Update

  • Force update from local source:

    uv tool install . --force
    
  • Upgrade to the latest version via PyPI:

    uv tool upgrade sari
    

Uninstall

uv tool uninstall sari

Project details


Release history Release notifications | RSS feed

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

sari-0.6.5.tar.gz (306.1 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

sari-0.6.5-py3-none-any.whl (299.7 kB view details)

Uploaded Python 3

File details

Details for the file sari-0.6.5.tar.gz.

File metadata

  • Download URL: sari-0.6.5.tar.gz
  • Upload date:
  • Size: 306.1 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for sari-0.6.5.tar.gz
Algorithm Hash digest
SHA256 756e2dd052924dac172256d6d3c42aa9eb09083859c8a151b4775a093296a4a5
MD5 df0c3133c8029a3892736f5efdbb036a
BLAKE2b-256 1cf96617066a1485ca2d25e0e7cbfc8b0835f9b662844d109e3d2127ff5bdb7e

See more details on using hashes here.

File details

Details for the file sari-0.6.5-py3-none-any.whl.

File metadata

  • Download URL: sari-0.6.5-py3-none-any.whl
  • Upload date:
  • Size: 299.7 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for sari-0.6.5-py3-none-any.whl
Algorithm Hash digest
SHA256 4988218b57f61f5e6b07016e2923a7d8dc9ba78622cd8e06b3bcb40a06aba62d
MD5 18367e2a59daf5869dfbd81a60f9aaf2
BLAKE2b-256 dd09d8b0ca59f2f7798ac5eed4a6c15ff54c1622fab261ddd5681701b45882d2

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page