OpenCoder AI Agent
An AI Super agent for VS Code that works with any model. Hand it a task and it reads your project, plans, edits files, and runs terminal commands — showing every action as it works.
Use it with OpenRouter, OpenAI, Anthropic, Gemini, your own API keys, or local models.
Features
- Multi-provider AI — OpenRouter, OpenAI, Anthropic, Gemini, xAI Grok, Microsoft Azure, AWS Bedrock, GCP Vertex AI, and any OpenAI-compatible endpoint (Ollama, LM Studio, vLLM)
- Agentic tool execution — reads/writes files, runs terminal commands, fetches URLs, inspects git diffs
- Subagents — delegate subtasks to specialized subagents that run in parallel and report back
- Agent swarm — an orchestrator agent coordinates multiple worker agents across complex, multi-file tasks
- Goal-based autonomous agents — give it a goal and it self-directs: planning, executing, and verifying until the goal is met
- Conversation history — every chat is saved and listed in the Activity Bar sidebar
- File & folder attachments — attach context directly to a message
- @ mentions —
@file.js injects file contents, @selection injects the selected text
- Plan mode — the agent drafts a step-by-step plan for review before executing
- Diff review — view git diffs inline, accept/reject individual changes
- CoWork — autonomous multi-step task execution on your files
- Skills & MCP — built-in skills (debug, refactor, test, docs, review) plus MCP server management
- Local user profiles — each profile keeps its own providers and API keys
- Secure key storage — API keys are stored in your OS keychain (macOS Keychain / Windows Credential Manager), never written to disk
- AGENTS.md — project-level agent instructions are loaded automatically
- Auto-update checks — notifies when a new release is available on GitHub
Getting Started
1. Create your profile
On first launch the welcome page asks for a name. Profiles store providers, API keys, and preferences — all locally on your machine.
2. Add an AI provider
Open the chat panel and click Settings (or run OpenCoder: Open Settings):
- Pick a provider (e.g. OpenRouter)
- Paste your API key
- Optionally set a model (e.g.
deepseek/deepseek-v4.1-flash-1017)
- Click Save & Test
Your key is stored in the system keychain — it is never written to settings.json or sent anywhere except the provider you choose.
3. Start coding
Open the chat from the Activity Bar (OpenCoder icon) and describe a task in plain English:
create a readme.md with install instructions
refactor src/utils.js to use async/await
run the tests and fix any failures
The agent will read files, make edits, run commands, and report results in the chat.
Layout
| Location |
Content |
| Left Activity Bar |
Conversation history + Quick Actions |
| Right / Secondary Sidebar |
Chat panel |
| Status Bar |
Quick-open launcher |
Click a conversation in the left panel to load it in the chat. Use + New Chat (tree title bar) to start fresh.
Slash Commands
| Command |
Action |
/terminal <cmd> |
Execute a shell command |
/read <path> |
Read file contents |
/write <path> <content> |
Write a file |
/dir <path> |
List a directory |
/editor |
Get active editor text |
/workspace |
Show workspace root |
/web <url> |
Fetch URL content |
/diff |
Show git diff |
/new |
Start a new conversation |
/export |
Export conversation to file |
/settings |
Open the settings panel |
/help |
Show help |
Or just type a question in plain English — no commands required.
VS Code Commands
| Command |
Description |
OpenCoder: Open Chat |
Open the chat panel |
OpenCoder: Open Settings |
Configure providers and preferences |
OpenCoder: New Chat |
Start a new conversation |
OpenCoder: Switch Profile |
Switch user profile |
OpenCoder: Plan Mode |
Plan a task before executing |
OpenCoder: Show Git Diff |
View uncommitted changes |
OpenCoder: CoWork Autonomous Task |
Run a multi-step task |
OpenCoder: Check for Updates |
Check GitHub for new releases |
Settings
| Setting |
Default |
Description |
openCoder.providerType |
— |
Provider: openrouter, openai, anthropic, gemini, grok, azure, aws-bedrock, gcp-vertex, openai-compatible |
openCoder.apiKey |
— |
API key (stored securely in keychain) |
openCoder.model |
— |
Model name (e.g. openai/gpt-4o) |
openCoder.apiBaseUrl |
— |
Custom base URL for OpenAI-compatible providers |
openCoder.defaultPermissionMode |
"" |
normal, acceptEdits, planMode, auto, yolo |
openCoder.submitWhileRunning |
queue |
queue or steer new messages |
openCoder.autoLoadAgentsMd |
true |
Auto-load AGENTS.md project instructions |
openCoder.autoCheckUpdates |
true |
Check GitHub for updates on startup |
openCoder.showStatusBar |
true |
Show the status bar launcher |
openCoder.maxFileSize |
1048576 |
Max file size (bytes) for read operations |
Security
- API keys are stored in the OS keychain via VS Code's SecretStorage — never in plain files
- Keys are only sent to the provider you configure
- All user profiles and preferences are stored locally
- No telemetry, no training on your code
Requirements
- VS Code
^1.85.0
- An AI provider account (OpenRouter, OpenAI, Anthropic, etc.) or a local model server
License
MIT — see LICENSE.
Links
| |
