Docs

User Guide

A practical guide for teams using Schift through the dashboard, API, SDK, CLI, and MCP surface.

Schift is a managed RAG infrastructure product. You upload documents, search or chat against them, and integrate the same data surface into apps through the API, SDK, CLI, or MCP server.

Choose the right surface

Surface	Use it when	Entry point
Dashboard	You want to create API keys, manage billing, and inspect buckets in a browser	`https://schift.io/app`
REST API	You need the thinnest integration from your own backend or service	`https://api.schift.io`
Python SDK	You want fast scripting, ingestion, search, or migration workflows	`pip install schift`
TypeScript SDK	You are integrating Schift from a TS or Node environment	`npm install @schift-io/sdk`
CLI	You want repeatable terminal commands and ops-friendly workflows	`npm install -g @schift-io/cli` → `schift`
MCP	You want ChatGPT or another MCP client to search Schift directly	`https://mcp.schift.io/mcp`

Typical flow

Create or copy an API key from the dashboard.
Upload files into a bucket or write documents into a hosted collection.
Use bucket search, query, or RAG chat depending on the product flow you need.
Wire the same key into your SDK, CLI, or MCP setup so local tooling and production use the same account.

Integration checklist

bash# REST API
export SCHIFT_API_KEY=sch_your_key_here

# Python SDK
pip install schift

# TypeScript SDK
npm install @schift-io/sdk

# CLI
npm install -g @schift-io/cli
schift --help

# MCP
# Remote MCP server URL: https://mcp.schift.io/mcp
# Authorization header: Bearer sch_your_key_here

Keep one source of truth

Use one live API key per environment and keep SCHIFT_API_KEY in the environment whenever possible. That lets the SDK, CLI, and MCP surfaces stay aligned without extra local config drift.

Most common operations

Goal	Recommended path
Upload raw files and let Schift process OCR, chunking, and embeddings	`POST /v2/buckets/{id}/documents` or the bucket helpers in the SDK
Run answer-ready knowledge search across uploaded documents	`POST /v2/buckets/{id}/search`
Return an answer with source citations in one call	`POST /v1/chat`
Use Schift as an OpenAI-compatible LLM router	`POST /v1/chat/completions`