indexed-sh 0.0.8

Index everything. Code, docs, and knowledge — one MCP, zero cloud.

Quickstart · Documentation · Blog & Guides

Local-first semantic search for AI agents. Give Claude Code, Cursor, Codex, and other MCP-compatible agents deep context over your codebase, docs, Jira, and Confluence.

Key Features:

• Privacy-first — all processing and storage happens locally, no data sent to third parties
• Semantic search — understands meaning, not just keywords, using dense vector embeddings
• Multiple sources — index local files (25+ formats), Jira, and Confluence (Cloud & Server)
• MCP integration — works with Claude Code, Cursor, Windsurf, Cline, and other MCP-compatible agents
• Incremental updates — keep your index fresh with git-based change tracking

Quick Start

# 1. Install indexed globally
uv tool install indexed-sh

# 2. Create a collection from local files
indexed index create files --collection my-project --path ./src

# 3. Search your collection
indexed index search "your query"

Why Indexed?

AI agents search your codebase on demand with grep — fast for small repos, but expensive on large ones. Every file read costs tokens. Every broad search burns context window.

Indexed pre-computes a semantic search index over your code, docs, and project tools, then exposes it via MCP. The result: instant, relevant context retrieval without the token overhead.

• Not just code. Index Markdown, PDFs, DOCX, PPTX, images, and 25+ formats via Docling. AST-aware code chunking via tree-sitter.
• Not just local files. Native Jira and Confluence connectors pull tickets, pages, and metadata into your index.
• Not cloud-dependent. Runs entirely on your machine. Local embedding models, FAISS for vector storage. No API keys required.
• Not one-shot. Incremental updates via indexed index update keep your index fresh as your codebase evolves.

MCP Integration

Indexed exposes a Model Context Protocol server for AI agent integration.

Claude Code (CLI, VS Code, JetBrains)

Works across all Claude Code surfaces — CLI, VS Code extension, and JetBrains plugin:

claude mcp add indexed -- indexed mcp run

This adds indexed to your project's .mcp.json. You can also manage servers via /mcp in the VS Code chat panel.

To add manually, create or edit .mcp.json in your project root:

{
  "mcpServers": {
    "indexed": {
      "command": "indexed",
      "args": ["mcp", "run"]
    }
  }
}

Cursor / Windsurf / Other Agents

Add to your agent's MCP configuration file:

{
  "mcpServers": {
    "indexed": {
      "command": "indexed",
      "args": ["mcp", "run"]
    }
  }
}

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "indexed": {
      "command": "indexed",
      "args": ["mcp", "run"]
    }
  }
}

Server Mode

# HTTP server mode
indexed mcp run --transport http --port 8000

Supported Sources

Installation

Requires Python 3.11+ and uv.

uv tool install indexed-sh

This installs indexed as a global CLI tool from PyPI.

uvx indexed-sh index search "your query"

git clone https://github.com/LennardZuendorf/indexed.git
cd indexed
uv sync
uv run indexed --help

Usage

# Create collections
indexed index create files --collection my-project --path ./src
indexed index create jiraCloud --collection jira-issues
indexed index create confluenceCloud --collection wiki

# Search
indexed index search "authentication flow"
indexed index search "bug reports" --collection jira-issues

# Manage collections
indexed index inspect                    # list all collections
indexed index inspect my-project         # inspect specific collection
indexed index update my-project          # update a collection
indexed index remove my-project          # delete a collection

# Configuration
indexed config inspect                   # view config
indexed config set search.max_docs 20    # set a value

For the full CLI reference and configuration guide, see the documentation.

Documentation

• Full Documentation — CLI reference, configuration, and guides
• Blog & Guides — tutorials and use cases
• Issue Tracker — report bugs or request features

Contributing

git clone https://github.com/LennardZuendorf/indexed.git
cd indexed
uv sync --all-groups
uv run pytest -q

Performance benchmarks (tests/benchmarks/, tests/system/) run in CI via pytest-bench-action — see .github/workflows/python-benchmark.yml. Run them locally with:

uv run pytest tests/system tests/benchmarks --benchmark-only

See the internal package docs and CLAUDE.md for architecture details.

License

See LICENSE file for details.

Credits

The Core v1 implementation is based on documents-vector-search by shnax0210, licensed under MIT and modified extensively.

indexed.sh · Star the repo if you find it useful!