Augment Code capabilities

A terminal-based agentic CLI that brings Augment's agent and Context Engine to the command line for interactive work and for headless automation.

An AI agent that completes software development tasks end to end: it breaks a request into a plan, then creates, edits, and deletes code across the workspace while keeping the codebase conventions.

An automated reviewer that reads a pull request with full-codebase context and posts prioritized inline comments on bugs, security issues, correctness, and cross-system breaking changes.

An in-editor chat that answers questions about the codebase, plans projects, and defines changes, drawing on full-codebase context before the agent implements anything.

In-editor code completions plus 'Next Edit', which propagates a change across the codebase by suggesting the related follow-on edits (e.g. updating all call sites, removing unused imports).

A proprietary retrieval system that semantically indexes and maps an entire codebase so the agent retrieves only the relevant slice for a task, rather than relying on grep or keyword matching.

A team-level platform that coordinates many coding agents across the SDLC with shared context and memory, running agents anywhere and keeping humans in the loop where judgment matters.

Native editor plugins that bring Augment's agent and chat into VS Code and the JetBrains IDE family.

Lets the agent call external tools and services — via Model Context Protocol servers and built-in native integrations — so it can read issues, query data, and act beyond the editor.

Lets users choose which underlying frontier model powers the agent, spanning Anthropic, OpenAI, Google, and Moonshot families, with per-model credit costs.

Four credit-based plans — Indie, Standard, Max, and Enterprise — where usage is metered in credits pooled at the team level.

Instruction files that steer the agent's behavior with your coding standards, conventions, and workflows, scoped per user or per project.