Give coding agents (Copilot Chat, Claude Code, Cursor, anything that speaks MCP — or anything that can shell out) live access to everything that lives inside VS Code and third-party extensions.
Explore the code with the tools the editor already has: Find References, Go-to-Definition, hover types, workspace diagnostics, git blame, symbol outlines, and any command a third-party extension registers.
Modify the code through the same language server that powers the editor:
structural rename, quick-fixes and refactors, organize imports, safe file
create/rename/delete via WorkspaceEdit, formatter output, source actions. Edits
go through the language server — safer than sed/awk on files because they
respect scope, imports, and semantics.
Implant exposes a single MCP tool, run_vscode_script, that executes a JavaScript
snippet inside the running VS Code extension host with full access to the
vscode.* API. Nothing runs without the user approving the exact snippet.
Quick start
- Install the extension.
- Open your project. Implant will offer to install itself into the workspace — this drops portable rule files that teach agents when to reach for Implant.
- Your agent picks up the tool automatically — MCP-native clients (Copilot Chat and others on VS Code ≥ 1.101) discover it via the registry, and CLI agents pick it up from the workspace instructions installed in step 2.
- When an agent invokes the tool, a webview opens beside the editor showing the script with syntax highlighting. Click Run, Run + trust session, or Deny.
Palette commands
| Command | Purpose |
|---|---|
| Implant: Install into Workspace | Writes five portable, safe-to-commit files: .vscode/implant/call.js (stdio↔HTTP MCP proxy), .mcp.json (Claude Code / generic MCP config), .cursor/mcp.json (Cursor MCP config), .claude/skills/implant/SKILL.md (Claude Code skill), .cursor/rules/implant.mdc (Cursor rules). Also upserts an <!-- BEGIN Implant --> … <!-- END Implant --> section into .github/copilot-instructions.md. Prompts before overwriting; asks which folder in multi-root workspaces. |
| Implant: Copy Agent Instructions | Copies the same instructions to the clipboard — paste into AGENTS.md, an agent's system prompt, or wherever your setup expects. |
Settings
| key | default | meaning |
|---|---|---|
implant.requireConfirmation |
true |
Show the approval webview before each script runs. |
implant.timeoutMs |
0 |
Hard execution timeout in ms. 0 disables it — long-running async scripts are cancelled by the user via a notification that appears after 10 seconds. Set to a positive value to enforce an automatic upper bound. |
implant.port |
0 |
Port for the HTTP MCP server. 0 picks an ephemeral one; the actual port is written to .vscode/implant/session.yml. |
The approval webview has a Run + trust session button that skips future prompts until the window is reloaded.
Safety notes
- Snippets run in the VS Code extension host with the same privileges as the extension itself. There is no sandbox — that is the entire point.
- The approval webview exists so you can eyeball each snippet before it runs. Turn confirmation off only when the agent is already in a tight review loop.
- The HTTP server only binds to
127.0.0.1and requires a per-session bearer token — the token lives in.vscode/implant/session.yml(mode0600, gitignored, regenerated on every activation). Requests without a valid token get401and raise a warning notification so you notice anything poking at the port. - Even with the token gate, don't run this on a shared machine: any process running as your user can read the session file.
- The confirmation prompt is the last line of defense. Only turn
implant.requireConfirmationoff when the agent is already in a tight review loop that you trust.