Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@space-ngrams search for the function 'calculateTotal' in the codebase"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
MCP server that gives AI agents superpowers to search through your codebase at lightning speed.
Space-nGrams connects to AI coding assistants (Claude Code, Codex CLI, Qwen CLI, OpenCode) and provides them with three essential tools: search code, find files, and read files. All powered by ripgrep for millisecond-level performance.
┌─────────────────┐ ┌──────────────────┐ ┌─────────────────┐
│ AI Agent │ │ Space-nGrams │ │ ripgrep │
│ (Claude Code, │────▶│ MCP Server │────▶│ (rg) │
│ Qwen, etc.) │◀────│ (Python) │◀────│ Search Engine │
└─────────────────┘ └──────────────────┘ └─────────────────┘
│ │ │
│ ▼ │
│ ┌──────────────────┐ │
│ │ Cache Layer │ │
│ │ (~/.space-ngrams│ │
│ │ /cache/) │ │
│ └──────────────────┘ │
│ │ │
│ ▼ │
│ ┌──────────────────┐ │
└─────────────▶│ Metrics & Logs │◀──────────────┘
│ (~/.space-ngrams│
│ /server.log) │
└──────────────────┘MCP (Model Context Protocol) is a standard that allows AI agents to access external tools and data sources. Instead of embedding all your code into the AI's context (which is slow and expensive), Space-nGrams gives the AI the ability to:
Search on demand — Find any pattern, function, or string in your codebase in milliseconds
Navigate efficiently — Locate files by name, then read only what's needed
Work with large codebases — No need to load everything into context
Blazing fast — SIMD acceleration, regex compilation, parallel search
Smart filtering — Respects .gitignore, skips binary files
Rich output — JSON format with line numbers and context
Battle tested — Used by developers worldwide daily
Repeated searches are common when AI agents explore code. Our cache layer:
Stores results for 5 minutes (configurable)
Persists across sessions on disk
Reduces latency from ~100ms to ~2ms on cache hits
Automatically manages size limits (50 MB default)
Related MCP server: codeweave-mcp
Feature | Benefit |
Caching | 10-50x faster for repeated searches |
Metrics | Track performance and cache hit rates |
Configuration | Customize limits, timeouts, ignore patterns |
Logging | Debug issues via |
Tool | Description |
| Search for regex/string in code with context |
| Find files by glob pattern |
| Read file content (max 200 lines/call) |
| Get session performance statistics |
git clone https://github.com/your-username/space-ngrams.git
cd space-ngrams# Windows
winget install BurntSushi.ripgrep.MSVC
# macOS
brew install ripgrep
# Linux
sudo apt install ripgrepVerify: rg --version
pip install mcppython src/server.pyThe process will wait on stdin — that's correct. Stop with Ctrl+C.
claude mcp add space-ngrams -- python /path/to/space-ngrams/src/server.pyVerify:
claude mcp list
# space-ngrams: python ... - ✓ ConnectedAdd to ~/.codex/config.toml:
[mcp_servers.space-ngrams]
command = "python"
args = [ "/path/to/space-ngrams/src/server.py" ]Add to ~/.qwen/settings.json:
{
"mcpServers": {
"space-ngrams": {
"command": "python",
"args": ["/path/to/space-ngrams/src/server.py"]
}
}
}Add to opencode.json in your project root or home directory:
{
"mcp": {
"space-ngrams": {
"type": "local",
"command": ["python", "/path/to/space-ngrams/src/server.py"]
}
}
}The AI agent automatically uses these tools when needed. You can also trigger them explicitly:
find all calls to getUserById in D:/Projects/MyAppshow all .ts files in the src folderread D:/Projects/MyApp/src/auth/service.ts lines 50-100Parameter | Required | Description |
| yes | Regex or literal string |
| yes | Directory or file to search in |
| no | File type filter, e.g. |
| no | Lines of context (default: 2) |
Parameter | Required | Description |
| yes | Glob, e.g. |
| yes | Root directory to search in |
| no | Maximum results (default: 100) |
Parameter | Required | Description |
| yes | Absolute or relative path |
| no | First line to read (default: 1) |
| no | Last line (inclusive) |
Returns performance statistics:
{
"search_code": {
"total_calls": 15,
"cache_hits": 8,
"cache_hit_rate": 53.3,
"avg_duration_ms": 45.2,
"min_duration_ms": 2.1,
"max_duration_ms": 234.5
}
}Create a config file at ./space-ngrams.toml (project-specific) or ~/.space-ngrams/config.toml (global):
[cache]
enabled = true
ttl_seconds = 300 # 5 minutes
max_size_mb = 50
[limits]
max_search_results = 50
max_files_results = 100
max_read_lines = 200
default_context_lines = 2
search_timeout_seconds = 15
[ignore_patterns]
# Additional patterns to ignore (beyond .gitignore)
ignore_patterns = [
"*.log",
"*.tmp",
"node_modules/**",
"__pycache__/**",
]
[metrics]
enabled = trueSee space-ngrams.example.toml for a full example with comments.
space-ngrams/
├── src/
│ └── server.py # MCP server implementation
├── pyproject.toml # Python package metadata
├── space-ngrams.example.toml # Example configuration
├── LICENSE
├── README.md # This file (English)
├── README_RU.md # Russian translation
└── docs/
└── ARCHITECTURE.md # Architecture notes (optional)📖 Also available in Russian.
Location | Purpose |
| Server logs and metrics |
| Persistent cache storage |
MIT
This server cannot be installed
Unclaimed servers have limited discoverability.
If you are the server author, to access and configure the admin panel.
An MCP server that gives your AI access to the source code and docs of all public github repos
Driflyte MCP server which lets AI assistants query topic-specific knowledge from web and GitHub.
Search your AI chat history (ChatGPT, Claude, Codex) from any MCP client. Remote, private, read-only
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/dev993848/mcp-fast-search-code'
If you have feedback or need assistance with the MCP directory API, please join our Discord server