Changelog

We’re thrilled to announce the rollout of Claude Sonnet 4.5, Anthropic’s latest and most capable coding model. Starting today, Sonnet 4.5 is the default model for Augment Code, and will appear alongside GPT-5 in the model picker. Sonnet 4.5 delivers faster, more collaborative code generation and editing across all of our plans. Sonnet 4 will continue to be available for a limited time.

Why we’re excited (and why you should be too)

Sonnet 4.5 makes smarter use of the codebase context provided by our Context Engine and is more steerable overall. When uncertain, it asks clarifying questions instead of heading down rabbit holes. It also persists state more effectively, handles longer-horizon tasks, and introduces multi-agent capabilities that open up new possibilities we’re actively exploring.

On our internal eval, which measures the agent’s one shot instruction-to-PR capabilities, Sonnet 4.5 scored nearly the same as Sonnet 4 in terms of quality.

In practice, it’s much more efficient:

• 34% fewer tool calls on average
• ~26% faster overall task completion time

The result: the same accuracy, delivered with more speed and flow.

What it means for you

• Smoother collaboration. Sonnet 4.5 checks its reasoning with you more often.
• Longer horizons. It handles longer-running tasks more reliably.
• Better rule adherence. Our early tests show that Sonnet 4.5 has better compliance with rules files.

We’ve observed that Sonnet 4.5 makes fewer assumptions and asks clarifying questions more frequently before proceeding. This behavior has been seen as positive overall in our testing. In contrast, Sonnet 4 tends to be more eager to complete tasks without the developer’s involvement.

We have updated our prompt for the Auggie CLI to ensure that Sonnet 4.5’s tendency to ask for clarification does not break automated workflows.

Model behavior can be adjusted to your preferences by using the rules files. For example:

• To make Sonnet 4.5 run more autonomously, add rules that reduce its clarifying questions or encourage faster action.
• To make Sonnet 4 less eager, add rules that require it to confirm assumptions or ask clarifying questions before proceeding.

How to get it

Sonnet 4.5 is our new default. That means if you are on the latest version of the Augment Code extension in VS Code or JetBrains, or using the CLI, new requests will use Sonnet 4.5 automatically as we roll it out to all customers over the next 24 hours.

To make sure you can access Sonnet 4.5 as fast as possible, we suggest restarting your extension.

Models will not switch mid conversation, and if you have previously chosen GPT-5 in the model picker, your requests will not switch to a different model. You can change this as desired by selecting your model of choice in the model picker.

Agents aren’t just chat anymore—they’ve grown into full systems that make tool calls, manage tasks, edit files, and update their own memories. As agents scale to handle more complexity, their interfaces need to scale too.

We’ve redesigned the experience with Agent Tabs:

• Tabs for Thread, Tasks, and Edits now live in the chat header.
• By using the full vertical space, each panel gives you both maximum readability and a dedicated view that helps you stay focused on the stage of the workflow you’re in.
• Cleaner ergonomics for navigating multi-step agent workflows.

This update trades horizontal header space for dramatically better visibility, so you can work with your agent like the full-fledged system it has become.

Available in VS Code v0.549.1+. IntelliJ support coming soon!

Securely pass secrets to your remote agent environments.

Remote agents often need secrets — like API keys or tokens — to build and run code in a fully functional dev environment. Until now, there wasn’t a convenient or secure way to inject them.

Remote Agent Secrets fixes this by letting you store and manage secrets directly in Augment, and have them automatically injected into your remote agent’s execution environment.

How it works

• Add required secrets in Augment’s secrets manager
• Configure your remote agent setup script to use those secrets (as env vars or files)
• Remote agents securely access them when running your code

Use cases

• Build processes requiring access to internal infrastructure
• Running or testing code that depends on cloud API keys
• Any workflow where secrets are required to run locally, now extended to remote agents

Availability

Remote Agent Secrets is available now in VS Code v0.555.1+.

Run multiple tools at the same time for faster agent execution.

Until now, agents ran tools one after another — generate, run tool, generate, run tool — adding latency to complex tasks and sometimes requiring extra turns to finish a job.

With Parallel Tool Calls, your agent can now execute multiple tools in one response, returning results faster and reducing overall wait time. Turns with Parallel Tool Calls run at least 2x faster than before, often more.

How it works

When the agent needs to run several tools, it will:

• Finish generating its plan
• Run tools 1 through n
• Return results in a single response

No configuration needed — the agent automatically decides when to call tools in parallel.

Use cases

• Tasklist updates run in the background without slowing down your turn
• Multi-file reads or edits complete in one step instead of many
• Faster codebase exploration with minimal latency

Availability

Available in VS Code v0.545.1+ and JetBrains v0.283.0+.

We’ve rolled out improvements to GPT-5 that make responses noticeably faster and more reliable, especially in multi-step agentic workflows.

What’s changed

• Reduced latency on GPT-5 responses
• Better continuity across tool calls
• Slight quality improvements in complex reasoning tasks

Why it matters

If you use GPT-5 with agents, you’ll notice:

• Faster execution of multi-tool workflows
• Responses that feel more intelligent and less fragmented
• A smoother overall experience in long-running sessions

Available in VS Code v0.545.1+ and JetBrains v0.283.0+

Curate and control the memories that your agent saves as you work together.

Memories help your agent stay aware across sessions, but sometimes unnecessary or low-quality memories can sneak in. Until now, the only way to audit them was by periodically opening the memories file.

With Agent Memory Review, you can now review, edit, and curate memories as they’re created — directly in the chat panel.

How it works

• When a new memory is created, you’ll see an “X Pending Memory” button in the turn summary
• Click it to review the new memory without leaving chat
• Approve, edit, or discard as needed

Use cases

• Power users who want to actively curate agent memories for accuracy
• Improving long-term memory quality for ongoing projects
• Reviewing memories in real-time to catch spurious entries early

Availability

Memory Review is available in VS Code v0.542.0+ and JetBrains v0.280.0+.

The following changelog covers changes from v0.5.0 to v0.5.2.

New Features

• Added /editor command to compose prompts in your preferred external editor (VS Code, vim, nano, etc.), use Ctrl+o for quick access
• Custom commands can now specify which AI model to use in their frontmatter configuration
• Slash commands can be run directly from the command line (e.g., auggie /help)
• Added ability to edit existing tool permission rules without recreating them (press e to edit)

Improvements

• Press ? to display more commands and shortcuts
• Improved paste functionality on Windows
• Improved support for IME input for Chinese and other languages
• Added slash commands to the history manager for easier command recall
• Added j/k keyboard navigation support for all menus (Task List, Model Picker, GitHub workflow)
• Display model names in footer when you select a non-default model

Changes

• Quiet mode (--quiet) now automatically defaults to text output mode
• Added workspace size limits to prevent indexing excessively large directories
• Prevented indexing when running CLI from home directory
• Automatically enter INSERT mode when typing `/vim` command
• Transcript defaults to compact output, use /verbose to toggle modes

Bugs

• CLI continues gracefully when settings fail to load instead of crashing
• Invalid MCP configuration no longer causes crashes
• Limited @ file cache size to prevent performance degradation when opened in large directories
• Fixed git worktree detection
• Fixed command exit behavior for command line – model list, tokens print, and session delete now properly exit after completion
• Fixed Ctrl+c and double escape shortcuts to work properly in vim mode

Improvements

• Made the edit tool more robust
• Stopped the terminal tool from getting stuck with invalid working directories
• Added support for AGENTS.md as a workspace guidelines source
• Improved chat performance and responsiveness
• Changed new chat shortcut to Cmd/Ctrl+N
• Improved the experimental terminal support
• Miscellaneous UI improvements

Bug Fixes

• Fixed task list editing where name and description changes were not persisting
• Fixed an issue where code blocks could not be applied
• Fixed integrations configuration in settings

Since launching in private beta a month ago, thousands of developers have unlocked the power of Augment in their terminal, using it as their daily driver coding agent and supercharging their teams by integrating Auggie in their development workflows.

Now, Auggie CLI is available to everyone, bringing Augment’s industry-leading context engine to every part of the software development stack.

What’s new

• Custom commands allow you to create specialized workflows and repeatable actions, like code review or updating docs. You can trigger commands on the command line with auggie command <name> or from slash commands in the terminal UI.
• All of your favorite features from the IDE are available in the CLI, like tasklist, prompt enhancer, and model picker.
• Tool permissions define exactly what tools an agent can use and how it can use them, ideal for using agents in automated systems like CI.
• Ready-to-use GitHub Actions to get Augment-powered code review and PR descriptions in minutes.

Augment everything

Software is not just built in an editor, it's built in your CI pipeline, ships through your deployment system, and runs in production. With auggie, you can bring our powerful agent and codebase intelligence to automatically fix test failures, run security assessments, and triage errors and alerts. Because auggie is a Unix-like utility complete with pipe support, it can run anywhere Node.js 22+ is available, including GitHub Actions and AWS Lambda.

Context matters

Deep codebase intelligence is the difference between a software agent that just works and one that makes you work. Augment’s context engine pulls in exactly the right information for the task automatically, no matter the size of your codebase.

Your terminal never felt so fresh.

Start using the Auggie CLI today with one command.

And we want to hear what you think, so we built a command for that: /feedback.

Happy coding!

New Features

• Added turn summaries to provide concise overviews of conversation context
• Added an experimental setting to control the implementation used for terminal support

Improvements

• Added a keyboard shortcut for ask a question mode
• Improved some tooltips
• Allowed using j/k keys to navigate up and down in menus
• Improved handling of MCP tools with long outputs
• Miscellaneous UI improvements

Bug Fixes

• Fixed Stripe integration
• Fixed issue preventing feedback forms from opening
• Fixed lag in chat input that was affecting typing performance
• Fixed rare issue causing menus to get stuck
• Fixed a regression preventing some common UI components from rendering
• Fixed an issue that caused the model picker not to open
• Fixed issue with closing dropdown menus
• Fixed a bug stopping the tools settings page from loading
• Fixed one cause of the "selected text exceeds the allowable limit" errors
• Fixed an issue with enhance prompt
• Fixed an issue that would cause the Augment panel not to load

Securely connect to SaaS-powered MCPs.

The MCP ecosystem is expanding — and many remote MCPs now require authentication. We’ve added built-in auth support so your agent can connect directly to these services.

How to Use

Setup is unchanged. If a remote MCP needs auth, you’ll see an Authenticate button and 🟠 status indicator. Click to connect.

Use Cases

• Automate setup: Configure connections or environment variables right from your agent.
• Speed up debugging: Pull logs from CircleCI, Sentry, or LaunchDarkly without leaving your workflow.
• Provision on demand: Create databases, tables, or cloud resources in Snowflake, Supabase, or Datadog.
• Integrate SaaS tools into tasks: Let your agent manage monitoring, deployment, or configuration steps end-to-end.

Availability

Available now for all VS Code users.

Imagine: you’re debugging a complex API issue with Augment.

Your agent examines logs across services, checks 12 files, tests endpoints, identifies a timing problem between microservices, and suggests fixes to 3 files—all in one turn.

Instead of scrolling through dozens of tool calls, the new Agent Turn Summary distills everything into a simple line—“32 tool calls, 3 files modified”—with expandable details. You see the full scope at a glance, and open the details only when you need them.

Review in seconds, not minutes.

How It Works

At the end of an agent response, you’ll now see a summary with:

• A summary and count of tool calls
• A quick snapshot of changes made

It appears alongside the feedback footer, so you can review activity right away without scrolling back through the entire turn.

Use Cases

• Reviewing long-running GPT-5 loops with 50–100 tool calls
• Spotting when an agent run is heavier or lighter than expected

Availability

Available in prerelease for VS Code and JetBrains.

At Augment, we noticed a common challenge: our team frequently needed quick information from the Agent without making any changes to the codebase. Whether asking for clarification on a change or reviewing a plan before implementation, we wanted a way to get answers without accidentally triggering modifications.

Quick Ask Mode is our solution to this exact problem. Get instant answers from the Agent — without making changes to your codebase.

When enabled, Quick Ask Mode restricts your agent to read-only tools, so it can explore and answer questions without modifying files. It keeps your full conversation context while giving you peace of mind and lets you easily switch back to the default Agent mode when you're ready to implement changes.

How It Works

• Toggle Quick Ask Mode from the Agent chat input
• Available in local agent mode only

Use Cases

• Planning: Analyze architecture, dependencies, and patterns to create a tasklist before coding
• Code understanding: “Explain how the authentication system works” — safe exploration with no edits
• Documentation queries: “What does this function do and where is it used?” — pull info from across the codebase
• Onboarding: Let new team members explore safely without risk of accidental changes

Availability

Quick Ask Mode is live for all users in VS Code and JetBrains.

New Features

• Added offline screen with troubleshooting tips when network connectivity issues are detected ## Improvements
• Added keyboard shortcut to open the context menu
• Added Escape key support to close full-screen overlays like the threads list
• Made the model selector responsive to prevent layout issues with small width windows
• Improved MCP server compatibility
• Improved styling of the expand/collapse buttons on tool blocks
• Miscellaneous UI improvements and bug fixes

Bug Fixes

• Fixed code block collapsing when copying content
• Fixed the Enter key to send while editing a message
• Fixed a rare issue that could cause the terminal tool to use the wrong shell
• Fixed one cause of the "selected text exceeds the allowable limit" errors
• Fixed MCP authentication infinite loop
• Fixed new thread shortcut also creating a new file

New Features

• Extended IDE Compatibility: Restored support for IntelliJ IDEA 2024.2, allowing users on older IDE versions to continue using Augment while maintaining compatibility with the latest 2025.x versions.
• Remote MCP OAuth Support: Added comprehensive OAuth authentication support for remote Model Context Protocol (MCP) servers, enabling secure connections to external services that require authentication.

Improvements

• UI Polish: Various UI improvements including better animation performance, fixed code block copy buttons showing proper "Copy" text, and improved tooltip displays in the history panel.
• Extension Status Dialog: The Extension Status dialog now displays the size of the webview state file, providing better visibility into storage usage and helping with troubleshooting.

Technical Improvements

• Improved conversation history persistence across sessions
• Better integration with the sidecar service for various operations

Improvements

• Improved context engine display by reducing frequency of "Show More" prompts
• Added keyboard shortcuts for cancel and resend message
• Added improved drag handle for resizing chat panels
• Improved chat rendering performance in conversations with many collapsed sections
• Improved the feedback panel
• Miscellaneous UI improvements

Bug Fixes

• Fixed one cause of the "selected text exceeds the allowable limit" errors
• Fixed issues with the terminal tool and the "exec" shell
• Fixed unwanted newlines when sending messages in chat
• Fixed feedback panel dropdown behavior
• Fixed regression on agent thread loads

Augment now supports two models in production — Claude Sonnet 4 (default) and OpenAI GPT-5 — with a new picker that lets you switch between them at any time. This is the first time we’ve exposed model selection directly in the product.

We’ve traditionally avoided surfacing model choice. But Sonnet and GPT-5 now handle different coding scenarios well enough to justify optionality:

• Use Sonnet for: fast answers, quick single-file edits, and iterative review loops
• Use GPT-5 for: multi-file changes, complex refactors, debugging, and edge-case handling

This lets you choose the right model based on the task at hand — without changing your workflow or context.

We’re also using switching patterns as signal: where users prefer one model over another helps guide future tuning, routing, and defaults.

🔗 Read the full announcement

Available in: JetBrains v0.265.2, VS Code v0.524.1. Rolling out gradually.

You can now use GPT-5 in Augment via the new model picker.

Available in: JetBrains v0.265.2, VS Code v0.524.1. Rolling out gradually.

To switch models, open the picker in your IDE sidebar and select GPT-5. Claude Sonnet 4 remains the default.

🔗 More on how the models compare

New Features

• Added OAuth support for remote MCP servers

Improvements

• Added keyboard shortcuts for enhanced prompt and tool approval
• Added loading spinner for rules when opening settings page
• Added tooltips to terminal commands in tool outputs to show full command text when truncated
• Enhanced thread navigation with keyboard support and auto-focus on selected threads
• Improved codeblock headers
• Improved the web fetch tool
• Improved code block styling and margins
• Added user email display in Settings → Account screen
• Miscellaneous UI improvements

Bug Fixes

• Fixed focus stealing issue where chat input would force focus on first load
• Prevented loss of message edits when clicking outside the editing box
• Fixed the share button
• Fixed text area auto-resize functionality
• Fixed an issue making it impossible to add/remove MCP servers

New Features

• Added a new UI for reporting bugs from within a conversation
• Redesigned threads list with improved filtering, search, and visual organization
• Added sound notification when tools require user approval

Improvements

• Added "New Thread" dropdown to the Threads panel
• New workspaces open agent mode rather than chat by default
• Fixed an issue where agent edits and autoformatters would conflict and cause edit loops
• Enhanced UI for MCP tool installations and calls
• Improved tool block expand/collapse design
• Improved thread navigation
• Added keyboard shortcut tooltips
• Made checkpoint headers clickable to toggle expansion of checkpoint content
• Made the agent less likely to kill running terminal processes
• Improved ability to navigate focus with tab/shift-tab
• Miscellaneous UI improvements

Bug Fixes

• Fixed an issue that could corrupt conversation history
• Fixed issues in line diff count component for code edits
• Fixed some errors that caused failures to launch terminal processes
• Fixed bug where toggling agent auto mode would incorrectly switch threads
• Disabled chat message navigation when the input field is focused
• Fixed an issue where the sign-in link did not work
• Fixed an issue where we would try to connect to disabled MCP servers on startup
• Fixed crashloops
• Fixed an issue that made it impossible to upload images
• Don't play sounds when skipping tool calls
• Fixed "New Thread" dropdown being blocked by tooltips

Since AI-powered coding assistants entered the scene, most have focused on code generation—helping you write a function here, debug a snippet there. But real software goes well beyond your editor: from local debugging to CI/CD pipelines, from GitHub Actions to custom deployment scripts.

Today, we're launching Auggie, an agentic CLI tool designed not just to be the most codebase-aware agent in your terminal, but to integrate into every part of your development workflow.

Augment where you work

Now Augment is available directly from your command-line with a rich terminal interface, Visual Studio Code, and JetBrains IDEs. Whether you prefer command-line workflows or integrated development environments, Augment works where you work.

Automation and integration ready

Auggie works as a unix-style utility, perfect for integration into your existing scripts and automation. Use it for code review in your CI pipeline, automated testing, or as part of your deployment process. Use the built-in `/gitub-workflow` command to quickly generate GitHub Actions for Pull Request descriptions and Pull Request reviews.

# Use it on your command line
auggie "Generate address test data formatted as json" > example.json

# Pipe data into Auggie
git diff | auggie "Explain the impact of these changes"

# Put it in your build system
auggie –print "Check if all new functions have tests"

Deep Codebase Intelligence

Auggie doesn't just see the files you show it—it knows and understands your entire codebase autonomously. Using Augment’s leading context engine, it maps project structure, dependencies, and patterns without manual context selection. Whether you're working on a small personal project or a million-line enterprise codebase? Auggie scales with you.

Getting Started

Auggie is rolling out today to Enterprise customers and will be available to teams and individuals over the next few weeks. We're releasing Auggie in beta, and we're actively gathering feedback from you to improve the experience and see what you can do with it. Tell us how Auggie can help do the things you need so you can focus on the things you love.

npm install -g @augmentcode/auggie

Just can't wait? Join the waitlist to get your hands on it early. If your company or team would like to try Auggie today, get in touch.

[\*o\*] Major Features

• Task List Integration: You can now access and manage your development tasks directly within the JetBrains environment, streamlining your workflow by keeping track of to-dos and project items without leaving your IDE

[-_-] Improvements & Fixes

• Smoother File Operations: File creation now happens seamlessly in the background, preventing version control dialogs from interrupting your chat flow
• Polished Chat Experience: Enjoy cleaner message layouts, better visual spacing, and more intuitive navigation throughout your conversations
• Enhanced Performance: Better memory usage, improved sidecar launching, and more reliable background processes contribute to a smoother plugin experience
• Chat Interface Enhancements: Fixed chat message edit box width issues and improved text wrapping on feedback panels for better readability
• Image Handling: You can now preview images in a modal window, and images are automatically inserted at the end of chat content for consistent formatting
• Keyboard Shortcuts: Improved keybinding infrastructure with the Cmd+. shortcut now cleanly toggling between auto and manual modes instead of cycling through all modes
• Tool Integration: Fixed missing play buttons for fetch and MCP tools, and improved tool component click behavior and hover styles

Tasklist helps your Agent break down complex problems into clear, actionable steps. You can review the plan, run tasks one by one, and add new ones as you go. It’s a simple way to keep track of progress—without losing context.

Getting Started

Your Agent will suggest a Tasklist when it spots a multi-step problem. You can also kick off a Tasklist yourself. Here’s how:

• Click the checklist icon next to Changes
• Add tasks with the ➕ button or type directly in the prompt bar
• Or just say: “Start a Tasklist to…”

How Tasklist Helps

• Break down open-ended problems into manageable steps
• Keep context by carrying your Tasklist across chats
• Revisit completed tasks or explore alternatives
• Stay organized by removing irrelevant steps
• Track progress with visual task states (not started, in progress, done)

A Few Things to Know

• The Agent can generate and nest subtasks automatically
• Run tasks individually—or all at once
• Stop and redirect any task mid-run
• Switch to Changes view to see what the Agent did

Works with Your Stack

Use Tasklist alongside Jira or Linear to break down tickets, track steps, and post updates. Or use it on its own to stay organized across threads.

New Features

• Added ask mode for agents - a question-only mode that gathers information without making any code changes
• Added native OAuth integration for Stripe and Sentry MCP tools
• Added Redis support to Easy MCP installation

Improvements

• Made entire header area clickable for expanding/collapsing tools
• Improved tool block UI
• Changed Cmd+. keyboard shortcut to toggle between auto and manual modes
• Added delete option to the remote agent header menu
• Added a confirmation dialog before modifying remote agent SSH configs
• Allow chat to use the full width of the panel
• Miscellaneous UI improvements

Bug Fixes

• Fixed unnecessary scrollbar appearing in empty chat conversations
• Fixed button layout issues when the chat panel is narrow
• Show a button to open the terminal after launching a process
• Made the agent less likely to kill running processes

Improvements

• Removed the auto mode warning dialog
• Improved mode switcher tooltips with clearer descriptions
• Added pagination support for remote agent repository selection to handle users with many repositories
• Fixed chat input collapsing unexpectedly when typing multi-line messages in remote agent mode
• Improved chat auto-scroll behavior
• Improved UI for generating and saving setup scripts for remote agents
• Improved thread selector dropdown UI
• Ensured images are inserted at the end of chat content
• Supported previewing images in the chat editor
• Miscellaneous UI improvements

Bug Fixes

• Improved chat input undo functionality
• Fixed an issue where we would open chat mode instead of agent mode
• Fixed an issue where the chat input box would not be full width

New Features

• Added better native diff view integrations for agent-generated edits
• Add conversation navigation controls
• Add a setting to play a sound when the agent completes a turn

Improvements

• Change copy behavior to copy the entire chat
• Reduce clutter in the footer below agent/chat responses
• Show individual MCP tools immediately after adding the server
• Ask for confirmation when applying remote agent changes requires stashing local changes
• Improved Remote Agent onboarding experience
• Enhanced diff decorations
• Improved selection chip tooltips with clickable file paths and better formatting
• Improved Rules Editor interface with better layout, clearer labels, and enhanced user experience
• Miscellaneous UI improvements

Bug Fixes

• Improved terminal support for symlinked shell paths
• Fixed issue where switching to chat mode would sometimes fail
• Fixed issue where quick fix actions would sometimes open in the wrong mode
• Fixed automatic creation of empty new chat conversations when restarting VSCode
• Fixed issues with renaming threads
• Fixed diff view descriptions for new files in remote agent
• Fixed merge conflict card display in remote agent diff view
• Fixed copy button to include full text response
• Fixed multi-repository dropdown menu

User Interface & Experience

• Chat History Navigation: Added up/down navigation buttons and keyboard shortcuts (Cmd/Ctrl + ↑/↓) to navigate between conversation messages for improved chat browsing experience
• Redesigned History View: Updated History interface with tabbed navigation for better organization
• MCP Setup Improvements: Streamlined MCP server configuration by combining remote MCP buttons and expanding MCP install section by default

Developer Experience

• MCP Tools: Enabled MCP tools display in settings panel with individual tool visibility and server name mapping

Bug Fixes

• Copy Response Fix: Resolved issue where copy response functionality only copied the last chat response block instead of entire history
• Guidelines Chip: Fixed disappearing guidelines chip in the user interface
• Sound Effects: Corrected sound effects to stop playing when agent conversation is cancelled or user is not in debug features

Augment Rules are instruction files that guide your AI agent's behavior, giving you granular control over how your AI agent behaves and responds to your specific needs. These rules help the agent understand your preferences, coding standards, project requirements, and specific workflows.

Getting Started

Create rules files in the .augment/rules folder in your project directory. You can name these files whatever you'd like and organize them however makes sense for your workflow. For example:

.augment/
└── rules/
    ├── coding-standards.md
    ├── project-guidelines.md
    └── review-checklist.md

Three Ways to Use Rules

1. Always - Rules are attached to every query automatically
2. Manual - You manually select which rules to include for each query
3. Auto - Describe what the rules does, and the agent will intelligently fetch relevant rules

Key Features

• Smart Rule Selection: With Agent Requested mode, simply describe your task and the agent will automatically determine which rules are most relevant to include in the context.
• Easy Migration: Already using rules from other tools? We support automatic import from competitors' rules and guidelines folders, making migration seamless.
• Multiple Access Points: Find Rules settings in the settings panel or create a .augment/rules folder to get started immediately

Migration Notice

If you're currently using .augment-guidelines.md, don't worry! Your existing setup will continue to work exactly as before. However, we encourage migrating to the new Rules system for enhanced flexibility and organization.

Improvements

• Added a hotkey for toggling between chat and agent modes
• Improved chat stream reliability
• Added a button to copy the request ID on user messages
• Don't show the terminal automatically when running commands
• Show individual MCP tools in the settings page
• Removed chat history size warning dialog
• Redesigned History view
• Added error notifications when attempting to attach unsupported file types

Bug Fixes

• Fixed an issue with scrolling in the chat window
• Fixed an issue where the guidelines chip disappeared