MisterPilot

An AI coding agent that lives in your VS Code sidebar. Ask questions, let it explore your codebase, edit files, and run commands — all without leaving your editor, and nothing happens on your machine without your say-so (unless you tell it otherwise).

Powered by Multiple AI providers — pick the model that fits the task from the model picker, and bring your own API key from any supported provider.

Features

Agent & Ask modes

• Agent mode — the full agent: reads your code, edits and creates files, runs terminal commands, and uses your MCP tools. It loops through model ↔ tool round trips until the task is done.
• Ask mode — read-only. The agent can explore and search your workspace but can't change anything; it suggests changes instead.

Understands your workspace

The agent has a built-in toolset that runs locally in the extension:

You stay in control

• Diff review — every edit is shown as a diff card with Apply, Reject, Always allow, and View full diff (opens VS Code's side-by-side diff editor).

• Terminal approval — commands appear as a card with Allow, Deny, and Always allow. Output, exit code, and duration show inline; commands are killed after a configurable timeout.

• Always allow (auto-approve) — skip the approval step for any of three actions, each switched on separately:

  • Edits to existing files
  • New files
  • Terminal commands

  Turn them on from the Always allow button on a card, or from the ⚙️ settings strip in the chat panel. A dot on ⚙️ tells you when anything is auto-approved, and auto-approved actions are tagged auto in the chat.

• Destructive-command guard — even with terminal auto-approve on, risky commands always ask first and are flagged with a warning. That includes rm -rf, rmdir /s, del /s /q, Remove-Item -Recurse, git push --force, git reset --hard, git clean -f, git branch -D, DROP/TRUNCATE TABLE, disk formatting, shutdown/reboot, sudo/runas, npm publish, piping curl/wget into a shell, and chmod -R 777.

Chat history

• Conversations are saved automatically per workspace (up to 100 chats) and restored when you reopen VS Code.
• Switch between chats, start a new one, rename, or delete from the chat list.

Context & cost

• Adjustable context window — 128K to 1M tokens, set from ⚙️. Older history and tool output are trimmed to stay inside it, and a usage meter shows how full the context is.
• Model picker — switch between models from any of the supported providers, per chat.
• Session cost — see what the current chat has cost (USD, plus INR where available).

MCP servers

Extend the agent with any Model Context Protocol server — same config format as Claude Desktop. Manage servers from the MCP page in the chat panel or run MisterPilot: Configure MCP Servers. Connection status is shown for each server, and MCP tools are available in Agent mode.

"misterpilot.mcpServers": {
  "filesystem": {
    "command": "npx",
    "args": ["-y", "@modelcontextprotocol/server-filesystem", "C:/projects"]
  }
}

Why multiple providers?

Most AI coding tools lock you into one model and one subscription. MisterPilot works with three AI providers, so you can:

• Use the provider you already have an API key for
• Pick a fast, cheap model for quick questions and a stronger one for bigger tasks
• Pay only for what you use — no expensive monthly subscription
• Avoid lock-in to a single vendor's pricing

Getting Started

Step 1 — Get an API key

1. Create an account with any of the supported AI providers
2. Generate a new API key from the provider's dashboard
3. Copy the key — you'll need it in the next step

Step 2 — Install and configure the extension

1. Install MisterPilot from the VS Code Marketplace
2. Open the MisterPilot panel from the activity bar (look for the icon on the left)
3. Open the command palette with Ctrl+Shift+P (or Cmd+Shift+P on Mac)
4. Run MisterPilot: Set API Key and paste your key
5. Start chatting

That's it — no other configuration needed.

Commands

Settings

How It Works

MisterPilot sends your messages to the MisterPilot engine (engine.misterpilot.online) — or any OpenAI-compatible endpoint you configure — which runs the model. The model decides which tools to call; the extension runs them. All file system and terminal operations execute locally in your VS Code extension — edits and commands only run after your approval, or automatically if you've turned on Always allow for that kind of action.

Privacy & Data Security

What leaves your machine:

• Your messages and the context the agent gathers — the files it reads, search results, diagnostics, and command output — are sent to the model endpoint so it can answer. The agent only reads what it needs for the task; it doesn't upload your whole workspace.
• Requests go to the MisterPilot engine (engine.misterpilot.online) by default. The engine applies PII filtering (emails, phone numbers, credentials, and similar identifiers) before content is forwarded to the model.
• If you set your own misterpilot.baseUrl, requests go directly to that endpoint and the MisterPilot engine's filtering does not apply.

What stays on your machine:

• Your API key is kept in VS Code's encrypted secret storage, never in settings files.
• Chat history is stored locally in VS Code's extension storage, per workspace.
• All file edits and terminal commands execute locally in the extension, not on a server.

Tip: avoid asking the agent to read files that hold secrets (.env, private keys), since their contents would be sent to the model.

Documentation

Full documentation, guides, and API reference at misterpilot.online/docs

Requirements

• VS Code 1.85 or newer
• An API key from one of the supported AI providers