Intent

Product Introduction

1. Definition: Intent by Augment Code is a purpose-built, agent-driven developer workspace and integrated development environment (IDE) designed to orchestrate multi-agent systems for autonomous software engineering. Unlike traditional text editors, Intent functions as a specialized execution layer where a team of AI agents—coordinated by a central intelligence—manages the full software development lifecycle (SDLC), including implementation, verification, and documentation within isolated, containerized environments.

2. Core Value Proposition: Intent exists to solve the "context gap" and "spec rot" prevalent in modern software engineering. By shifting the workflow from Code-First to Spec-Driven Development, Intent ensures that the technical requirements (the "Spec") remain a living, breathing source of truth that evolves alongside the codebase. It leverages a proprietary Context Engine to provide agents with repo-level semantic understanding, allowing for high-fidelity code generation and complex cross-service refactoring without manual developer intervention.

Main Features

1. Agent Orchestration and Hierarchical Coordination: Intent utilizes a multi-agent architecture where a "Coordinator Agent" acts as the project manager. Upon receiving a feature definition or spec, the Coordinator breaks down the task into sub-components and delegates them to "Specialist Agents" (e.g., Auth Agent, API Agent, Test Agent). These specialists run in parallel, executing tasks such as implementing JWT authentication using RS256 signing or updating API gateway middleware. A "Verifier Agent" then checks the output against the original spec to ensure compliance before merging.

2. Living Specs and Spec-Driven Development (SDD): At the heart of Intent is the "Living Spec." Unlike traditional Product Requirement Documents (PRDs) that become obsolete once the first line of code is written, Intent’s specs are dynamically updated as agents complete work. If requirements change, the updates propagate to all active agents instantly. This ensures that the documentation and the implementation are always in perfect alignment, eliminating the "what did we decide?" ambiguity in team settings.

3. Context-Rich Isolated Workspaces: Every project in Intent runs in an isolated workspace equipped with a built-in code editor, terminal, git client, and a Chrome-based browser for real-time previewing. These workspaces are powered by the Augment Context Engine, which indexes the entire codebase, including dependencies, recent changes, and documentation. This allows agents to understand deep semantic relationships between files (e.g., how a seat-usage.ts service interacts with a billing.service.ts) to prevent regressions during complex refactors.

4. Multi-Model Flexibility and Resumable Sessions: Intent supports various state-of-the-art Large Language Models (LLMs), allowing developers to assign specific models to specific tasks—such as using Claude 3.5 Opus for complex architecture planning and Sonnet for rapid code iteration. Additionally, the platform features resumable sessions; the workspace state, including terminal logs and agent progress, persists across sessions. Auto-commit and built-in branch management ensure that no progress is lost if the user closes the application.

Problems Solved

1. Context Fragmentation and Tool-Switching: Developers often lose productivity by switching between their IDE, terminal, git client, and browser, while simultaneously trying to feed context to an AI chat window. Intent consolidates these into a single window where the AI agents have native access to every tool, reducing cognitive load and manual "copy-pasting" of context.

2. Outdated Documentation (Spec Rot): In traditional environments, documentation is a secondary concern that quickly drifts from reality. Intent solves this by making the spec the operational center. If the code changes, the spec updates; if the spec changes, the agents update the code.

3. Target Audience: Intent is designed for senior software engineers, technical leads, and AI-forward development teams who are building complex, multi-service architectures. It is particularly valuable for developers managing distributed systems, microservices, or large-scale refactoring projects where manual coordination of tasks is time-consuming and error-prone.

4. Use Cases:

• Cross-Service Feature Implementation: Implementing a shared authentication logic (like JWT with refresh tokens) across an API gateway and multiple microservices simultaneously.
• Automated Refactoring: Migrating legacy codebases to new frameworks or languages while ensuring integration tests remain green.
• Rapid Prototyping from Specs: Converting a high-level technical requirement into a functional, verified PR with minimal manual coding.

Unique Advantages

1. Differentiation from AI Chat Extensions: While standard AI coding assistants (like Copilot or basic chat extensions) function as autocomplete or "chat-with-file" tools, Intent is a "workspace-as-an-agent." It does not just suggest code; it plans, executes, tests, and verifies entire features across multiple files and services autonomously.

2. Integration with Existing Subscriptions: A key innovation is Intent's "Bring Your Own Agent" model. Users can utilize their existing subscriptions for Claude Code, Codex, or OpenCode directly within the Intent workspace. This allows teams to explore spec-driven development and agent orchestration without necessarily switching their entire AI provider ecosystem.

3. Proprietary Context Engine: The depth of information provided to the agents—ranging from raw code to recent issues and dependencies—exceeds standard RAG (Retrieval-Augmented Generation) implementations found in most IDEs, enabling agents to handle "repo-level" tasks rather than just "function-level" tasks.

Frequently Asked Questions (FAQ)

1. Does Intent support Windows or Linux? Currently, Intent is optimized for macOS, specifically for Apple Silicon (M-series chips). While there are no immediate releases for Windows or Linux during the public beta, the development team is monitoring platform demand for future expansion.

2. How does Intent handle token usage and pricing? During the public beta, Intent consumes regular Augment credits. There is no separate subscription fee for the workspace itself. Users can also integrate their own third-party AI provider keys (like OpenAI or Anthropic) to power the agents within the workspace.

3. Is my code secure within Intent's isolated workspaces? Intent is built with enterprise-grade security in mind, featuring a dedicated Trust Center and compliance standards. The isolated workspaces ensure that agent activities are contained, and the platform provides full visibility from the first commit to the final merge, allowing developers to review every change made by the agents.

4. What models can I use with Intent? Intent supports a variety of state-of-the-art models, including Claude 3.5 Opus, Sonnet, and GPT-4 series. It allows for "per-task" model selection, ensuring you use the most cost-effective or highest-reasoning model depending on the complexity of the specific agent task.