Skip to main content

​
What is the Context Engine?

Every AI uses the same models. Context is the difference. Augment’s Context Engine maintains a live understanding of your entire stack:
  • Semantic understanding — Not just grep. A full search engine for code that understands meaning, not just text.
  • Relationship awareness — Understands how files connect across repos, services, and architectures.
  • More than code — Indexes commit history, codebase patterns, external sources (docs, tickets), and tribal knowledge.
  • Smart curation — Retrieves only what matters and compresses context without losing information.

​
Local vs Remote Server

AspectLocal ServerRemote Server
Best forMaking edits, new features, active developmentAdding to codebase, understanding/talking to code
What’s indexedWorking directory in real-timeSelected repos’ default branches (chosen during GitHub App install)
SetupRun Auggie CLI locallyGitHub App + app.augmentcode.com/mcp/configuration
  • Run the Auggie CLI locally as an MCP server (stdio)
  • Your client accesses codebase-retrieval for your working directory
  • Indexes in real-time as you edit — no manual sync required

​
Remote Server (Best for Adding/Understanding Code)

  • Connect to Augment-hosted Context Engine over HTTP
  • Integrates with Augment GitHub App
  • Best for:
    • Adding to an existing codebase
    • Talking to / understanding the codebase
    • Cross-repo context
    • CI/server environments without a local working tree

​
Credit cost for MCP

For a limited time, all users who use the Context Engine MCP in February will receive 1000 free context engine MCP tool calls.
Context Engine MCP consumes credits at different rates per query based on the underlying costs and amount of retrieval required. In our experience, we’ve seen the average per query cost between 40-70 credits per query.

​
How Indexing Works

​
Local Indexing

  • Real-time indexing of your working directory
  • Picks up local file changes immediately
  • No manual sync required

​
Remote Indexing

  • Repositories connected via the Augment GitHub App are automatically indexed
  • What’s indexed: All selected repositories’ default branches (selected during GitHub App installation)
  • When it updates: Automatically when commits are pushed to the default branch

​
Connect to MCP

  1. Sign up at augmentcode.com
  2. Get your configuration at app.augmentcode.com/mcp/configuration
The configuration page provides ready-to-use setup instructions for your MCP client, including authentication details for both interactive (OAuth) and non-interactive (API key) use cases.

​
Quickstart Guides

Get started with Context Engine MCP in your favorite AI tool:

Claude Code

Set up Context Engine MCP with Claude Code

Codex

Set up Context Engine MCP with OpenAI Codex CLI

Cursor

Set up Context Engine MCP with Cursor

Zed

Set up Context Engine MCP with Zed

GitHub Copilot

Set up Context Engine MCP with GitHub Copilot

OpenCode

Set up Context Engine MCP with OpenCode

Kilo Code

Set up Context Engine MCP with Kilo Code

Kiro

Set up Context Engine MCP with Kiro

AntiGravity

Set up Context Engine MCP with AntiGravity

Gemini CLI

Set up Context Engine MCP with Gemini CLI

Roo Code

Set up Context Engine MCP with Roo Code

Droid (Factory.AI)

Set up Context Engine MCP with Droid