A VS Code extension for AI-driven development where you control what the AI sees. No auto-indexing, no hidden magic. Works with LM Studio, Ollama, OpenAI, Groq, and more.
Switch between modes with Tab or click the badge. Each mode tailors the AI's behavior to your current workflow.
Just chat — no tool execution, no file changes. Quick Q&A, exploration, and brainstorming without side effects.
Read-onlyExplores and plans first, then shows the plan for your approval. When you confirm, it executes all changes.
Approval requiredFull auto-execution with tool chaining, undo support, diff preview, and self-healing. Maximum productivity.
Auto-executeBuilt for real development workflows — from quick questions to full sprint cycles.
Ghost text autocomplete as you type, powered by your local model. Tab to accept. No cloud dependency.
Paste screenshots and images into chat. Ask about UI designs, error messages, or diagrams.
Type @agent to invoke agents, #file to attach context, @skill to load best practices. Autocomplete on both.
Drag files from the VS Code explorer into chat. Automatically inserts #relative/path at cursor and attaches the file.
Load best-practice instructions with @skill react or @skill testing. Create your own as markdown files.
The AI can query definitions, references, hover info, diagnostics, code actions, and more — just like VS Code's IntelliSense.
After every write, the extension auto-fixes lint and compile errors using VS Code code actions. No manual fix cycles.
When token utilization reaches 60%, older messages are summarized automatically to prevent hallucination and keep context fresh.
Create AI personas with custom system prompts. Switch agents on the fly — the system prompt changes accordingly.
Full sprint workflow: plans, tasks, story points, dependencies, sprints, velocity tracking, and retrospectives.
Persistent memory across sessions — decisions, preferences, patterns. Keyword-retrieved and auto-injected into prompts.
Model Context Protocol integration with filesystem, terminal, and HTTP tools. Every action requires your approval.
Control context and invoke agents with simple text commands in the chat input.
All the commands at your fingertips. Configure them in VS Code keyboard shortcuts.
| Command | Keybinding | Description |
|---|---|---|
| Open Chat | Ctrl+Shift+A | Open the AI chat |
| Ask Selection | Ctrl+Shift+E | Ask about selected code |
| Ask (No Context) | Ctrl+Shift+M | Free-form question |
| Explain Code | Ctrl+Shift+H | Explain current file |
| Fix Errors | Ctrl+Shift+F | Fix diagnostics in file |
| Generate Tests | Ctrl+Shift+T | Generate unit tests |
| New Session | Ctrl+Shift+N | Save thread and start fresh |
| Stop Generation | Ctrl+Shift+X | Cancel current response |
| Inline Chat | Ctrl+K | Edit code inline |
| Auto-Fix | Ctrl+Shift+. | Auto-fix all errors |
| Retry | Ctrl+Shift+R | Retry last prompt |
Works with any OpenAI-compatible API. LM Studio is the recommended local option.
Download from lmstudio.ai, load a model, and start the local server.
Default: http://localhost:1234
Search "Explicit AI" in VS Code extensions or install from the marketplace.
Or use any OpenAI-compatible provider: Ollama, OpenAI, Groq, Together, etc.
Open the chat with Ctrl+Shift+A and start typing.
@planner Build a REST API for user management
@reviewer #src/auth/login.ts check this
How does #src/core/planner.ts work?
@skill reactClean separation of concerns. Every module has a single responsibility.
agents/ # Agent registry
chat/ # Chat session mgmt
commands/ # VS Code commands
completions/ # Inline completions
core/ # LLM, config, memory,
# planner, skills
mcp/ # MCP integration
threads/ # Thread persistence
tools/ # Workspace + LSP tools
ui/ # Webview & provider
extension.ts # Entry pointagents/ # Agent JSON configs
threads/ # Chat snapshots
skills/ # Markdown skill files
plans.json # Plans & sprints
memory.json # Project memoryTypeScript 5.3
VS Code API ^1.80
LM Studio / OpenAI API
No external AI deps
Dark mode native
No telemetry SDK