# Augment Code

> AI coding assistant with a Context Engine that understands your entire codebase, offering IDE agents, CLI, code review, and multi-agent workflows for professional software teams.

Augment Code is an AI-powered software development platform built around a proprietary Context Engine that maintains a live understanding of an entire codebase — including code, dependencies, architecture, and history. It targets professional engineering teams and enterprise organizations, offering agents that work across the IDE, terminal, and code review pipeline. The platform is actively developed and commercially available, with logos from companies including MongoDB, Spotify, Webflow, and others displayed on the vendor's homepage.

## What It Is

Augment Code positions itself as "The Software Agent Company," providing AI agents that span the full software development lifecycle — from writing code in the IDE to reviewing pull requests on GitHub. Unlike tools that rely solely on shared foundation models, Augment's differentiating claim is its Context Engine: a system that indexes and semantically understands a team's entire stack in real time, filtering thousands of sources down to the most relevant context for each task. The platform supports VS Code and JetBrains IDEs, a standalone CLI, and integrations with GitHub and Slack.

## The Context Engine Architecture

The Context Engine is the technical core of Augment Code. According to the vendor, it processes raw context — code, dependencies, documentation, style, recent changes, and issues — and applies semantic understanding to curate only the most relevant signals for each agent task. The homepage illustrates this with a figure showing 4,456 sources narrowed to 682 relevant ones. Augment publishes a benchmark comparing agent-generated pull requests against human-written code on the Elasticsearch repository (3.6M Java lines of code, 2,187 contributors), claiming its agent outperforms Cursor and Claude Code on overall quality, correctness, completeness, and best-practice adherence. The vendor also cites a SWE-Bench Pro leaderboard position of 51.80% for its "Auggie" agent, above Cursor (50.21%) and Claude Code (49.75%), attributing the advantage to context quality rather than model differences.

## Product Surface Area

Augment Code ships several distinct product surfaces:

- **IDE Agents** — Multi-step task lists, automatic memory across sessions, available in VS Code and JetBrains
- **Intent** — A macOS desktop workspace where a coordinator agent breaks work into parallel sub-agents, each operating in an isolated environment with a living spec
- **CLI (Auggie)** — A terminal-native interface with the same Context Engine and agent capabilities, no GUI required
- **Code Review** — AI-powered GitHub pull request review with inline comments, auto and manual modes, PR guidelines, and one-click IDE fixes
- **Slack Integration** — Available on enterprise plans for team-level coordination
- **Context Engine MCP** — An MCP (Model Context Protocol) server exposing Augment's context layer to other tools
- **Cosmos** — Described as "the operating system for agentic software," coordinating humans, agents, code, tools, policy, and memory at the org level; available on Max and Enterprise plans

## Deployment and Security Model

Augment Code is delivered as a cloud-connected service with IDE plugins for VS Code and JetBrains, a CLI, and a web app. The vendor states that no AI training is performed on customer data on any paid plan, enforced through Commercial Terms of Service. Security certifications listed include SOC 2 Type II, CMEK, ISO 42001 compliance, SIEM integration, data residency options, granular access controls, comprehensive audit trails, and enterprise SSO (OIDC/SCIM). A public Trust Center and Status Page are available.

## Update: Cosmos and Auggie Launch

The most recent notable product movement is the launch of Cosmos, described as an org-level operating system for agentic software that plugs into existing tools and compounds memory with every correction. Simultaneously, the vendor published SWE-Bench Pro results for its "Auggie" agent, claiming the top leaderboard position among named competitors. The changelog page is linked from the main navigation, indicating ongoing release activity. The blog and learn sections show posts dated through May 2026, confirming active product development.

## Features
- Context Engine with real-time codebase understanding
- IDE Agents for VS Code and JetBrains
- CLI agent (Auggie) for terminal-based workflows
- Intent: macOS desktop multi-agent workspace with living spec
- Cosmos: org-level agentic operating system
- AI-powered GitHub code review with inline comments
- Slack integration for team coordination
- Context Engine MCP server
- Automatic memory across agent sessions
- Task lists for complex multi-step work
- Credit-based usage with team-level pooling
- SOC 2 Type II certified
- CMEK and ISO 42001 compliance
- Enterprise SSO (OIDC/SCIM)
- No AI training on customer data
- Auto top-up credits
- PR summaries and inline comments
- One-click IDE fixes from code review

## Integrations
VS Code, JetBrains, GitHub, Slack, Jira, Linear, Notion, MCP (Model Context Protocol)

## Platforms
MACOS, WEB, API, VSC_EXTENSION, JETBRAINS_PLUGIN, CLI

## Pricing
Freemium — Free tier available with paid upgrades

## Version
1.0.0

## Links
- Website: https://www.augmentcode.com
- Documentation: https://docs.augmentcode.com/
- EveryDev.ai: https://www.everydev.ai/tools/augmentcode