# docsift

A local MCP server that converts documents to Markdown once, indexes them, and hands your agent only the passages that answer a question.

- MCP server · `type=mcp` · Claude Code, Codex, Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, Zed, Gemini CLI · ★ 7
- Author: anishmoncivarghese
- License: MIT · policy=link-only · description only; the file stays with the author
- Source: https://github.com/anishmoncivarghese/docsift
- Updated: 2026-08-12
- HTML: https://affpapa.org/en/nevibecode/mcp/docsift

## What it does

A 300-page PDF will not fit a model's context window, and pasting it in costs money anyway. DocSift converts the file once, splits it into token-budgeted chunks and writes everything to `./output/`. When the agent asks something, it gets back a few relevant chunks with page numbers and section headings instead of the whole document.

Two engines sit behind one interface: Docling for PDFs, MarkItDown for everything else. It runs on your machine — no cloud APIs, no accounts. Python 3.11 or newer is required. On Linux, install with `--torch-backend auto`, otherwise uv pulls the CUDA build of PyTorch and you download gigabytes of `nvidia-*` wheels your machine never loads.

Office support covers `.docx`, `.xlsx` and `.pptx` only. The pre-2007 binaries `.doc`, `.xls` and `.ppt` are refused by name, so save them in a modern format first. Browser clients like claude.ai and ChatGPT cannot use it: they need a remote transport, and DocSift speaks stdio only.

## Requirements
- Needs: Python and uv (commands run via uvx)

## Install

### Claude Code (native platform)

**Platform CLI**
```bash
claude mcp add docsift -- uvx docsift
```
uvx runs the command named after the package; if the README names another one, use uvx --from docsift <command>

### Codex

**Platform CLI**
```bash
codex mcp add docsift -- uvx docsift
```
uvx runs the command named after the package; if the README names another one, use uvx --from docsift <command>

### Cursor

**Config file**
File `.cursor/mcp.json`:
```json
{
    "mcpServers": {
        "docsift": {
            "command": "uvx",
            "args": [
                "docsift"
            ]
        }
    }
}
```
uvx runs the command named after the package; if the README names another one, use uvx --from docsift <command>

### GitHub Copilot

**Config file**
File `.vscode/mcp.json`:
```json
{
    "servers": {
        "docsift": {
            "type": "stdio",
            "command": "uvx",
            "args": [
                "docsift"
            ]
        }
    }
}
```
uvx runs the command named after the package; if the README names another one, use uvx --from docsift <command>

### Gemini CLI

**Platform CLI**
```bash
gemini mcp add docsift uvx docsift
```
uvx runs the command named after the package; if the README names another one, use uvx --from docsift <command>

## When to use it
- shrink a long PDF to a handful of relevant passages
- search your own files without uploading them anywhere
- get a citation with a page number and section heading
- turn `.docx` and `.pptx` files into clean Markdown
- split a document into chunks that fit a token budget

## Example prompts
- Find the revenue section in `report.pdf` and give me the page number
- Index `contract.docx` and tell me what the payment terms are
- Search my documents for anything about the Postgres integration

## Similar
- [REA](https://affpapa.org/en/nevibecode/mcp/rea.md) — MCP server · ★ 386989
- [GitHub](https://affpapa.org/en/nevibecode/mcp/github.md) — MCP server · ★ 33918
- [Netdata](https://affpapa.org/en/nevibecode/mcp/netdata.md) — MCP server · ★ 81115
- [browser-use](https://affpapa.org/en/nevibecode/mcp/browser-use.md) — MCP server · ★ 119160
- [Codebase Memory](https://affpapa.org/en/nevibecode/mcp/codebase-memory.md) — MCP server · ★ 47392
- [mcp-server](https://affpapa.org/en/nevibecode/mcp/mcp-server.md) — MCP server · ★ 43872

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md