Augment Code (Nightly)

Executes multi-file, multi-step coding tasks by leveraging a proprietary context engine that indexes and understands the entire codebase architecture, dependencies, and legacy patterns. The agent decomposes user intent into sequential edits across code, tests, and documentation, making decisions about which files to modify based on dependency graph analysis and architectural understanding rather than simple keyword matching.

Provides conversational Q&A interface where the LLM has access to indexed codebase context, allowing it to answer architectural questions, explain design patterns, and discuss implementation details with reference to actual code. The chat maintains conversation history and can reference specific files, functions, and dependencies discovered during codebase indexing.

Implements a guided editing mode called 'Next Edit' that suggests and executes sequential code modifications across multiple files (code, tests, documentation) in response to user direction. Rather than generating entire solutions at once, it breaks changes into discrete steps, allowing users to review and approve each modification before proceeding to the next coordinated edit.

Accepts natural language instructions to add, modify, or remove code across single or multiple files. The instruction engine parses user intent and generates appropriate code changes, leveraging codebase context to ensure modifications align with existing patterns, style, and architecture. Instructions can target specific functions, classes, or entire modules.

Generates real-time code suggestions as the user types, leveraging indexed codebase context to provide completions that align with project patterns, dependencies, and architectural conventions. Completions are triggered automatically or on-demand and consider multi-line context, function signatures, and imported modules to suggest relevant continuations.

Automatically indexes the workspace codebase to extract architectural information, dependency graphs, module relationships, and code patterns. The indexing engine supports 13+ programming languages and builds an internal representation of the codebase structure that powers all other capabilities. Indexing runs in the background and updates incrementally as files change.

Integrates with VS Code's extension API to provide access to Augment Code features through command palette commands, sidebar panels, and keyboard shortcuts. The extension hooks into VS Code's editor lifecycle to enable inline completions, context menus, and status bar indicators for agent status and indexing progress.

Supports code generation, completion, and analysis across 13+ programming languages (C, C#, C++, Go, Java, JavaScript, PHP, Python, Ruby, Rust, Swift, TypeScript, CSS, HTML) with language-specific context awareness. The system understands language-specific patterns, idioms, package managers, and build systems to generate contextually appropriate code.

Offers freemium pricing with free tier access to core features and premium tier for advanced capabilities. The nightly build provides early access to experimental features for users willing to accept instability in exchange for cutting-edge functionality. Pricing and feature tiers are not publicly documented in the marketplace listing.