PowerMem for VS Code

June 4, 2026 · View on GitHub

Give VS Code, Cursor, Windsurf, GitHub Copilot, and Qoder access to PowerMem intelligent memory. The recommended path is to let the current IDE's AI agent follow SETUP.md, which reuses or starts the right backend and configures only the current IDE/client unless you ask for a broader setup.

For Codex, OpenCode, and other generic MCP clients, use ../mcp-client/SETUP.md.

Features

Agent-guided setup — Paste one prompt in your IDE agent window and let it follow SETUP.md (Section 0 automates install, MCP, settings, and verification — no F5 or manual Setup wizard).
Seamless mode (default) — Linked AI tools (Cursor, Claude Code, Codex, etc.) retrieve memories via MCP automatically. No need to @powermem in chat.
Auto-capture on save — In seamless mode, matching files are added to memory when you save (configurable glob patterns and size limits).
Current-client linking — Configure the current IDE/client first; use broad linking only when you explicitly want every supported client config written.
Query memories — Search PowerMem from the editor (selection or query).
Add to memory — Save selection or a quick note to PowerMem.
@powermem chat participant — When seamless mode is off, use remember, save, and search commands in VS Code chat with auto-retrieve and periodic auto-summarize.
Dashboard — Quick access to query, quick note, and setup.
Health check — Status bar shows connection state; reconnect from the menu.

Requirements

VS Code 1.104+, Cursor, Qoder, or another VS Code–compatible IDE that can use PowerMem.
A configured PowerMem checkout or install. At minimum, set your LLM provider, API key, and model in .env or equivalent environment variables.
Backend priority:
1. Reuse an existing healthy HTTP API server if another IDE already started one: curl -s -m 5 http://localhost:8848/api/v1/system/health
2. Otherwise start the HTTP API server: powermem-server --host 0.0.0.0 --port 8848
3. Fall back to MCP-only only when the HTTP API cannot be made healthy: powermem-mcp streamable-http 8848, powermem-mcp sse 8848, or powermem-mcp stdio

Recommended Setup

First download the code and enter the directory:

git clone https://github.com/oceanbase/powermem
cd powermem

Copy .env.example to .env and set your LLM credentials, then open the AI agent window in your IDE and paste:

Read and follow apps/vscode-extension/SETUP.md to setup PowerMem

The agent-guided setup prefers powermem-server, reuses an existing healthy server when another IDE already started one, falls back to MCP-only only when HTTP API fails, and configures only the current IDE/client by default. After a fresh install, restart the IDE once so the extension and MCP config take effect.

Manual Quick Start

Use this only when you want to wire the extension by hand:

Install this extension in VS Code, Cursor, or Qoder.
Reuse or start powermem-server --host 0.0.0.0 --port 8848.
Click the PowerMem status bar item; if disconnected, run Setup and set Backend URL to the healthy HTTP API server, usually http://localhost:8848.
Configure only the current IDE/client. Use PowerMem: Link to AI Tools only when you explicitly want the extension to write all supported client configs.
Use Query memories or Add selection to memory from the command palette or status bar menu.

For a full setup checklist, see SETUP.md. For uninstall and cleanup, see UNINSTALL.md.

Seamless Mode vs @powermem Chat

Mode	Behavior
Seamless (default)	Linked AI tools get memory retrieval via MCP. File content is auto-captured on save (unless disabled). No `@powermem` needed.
Classic (`seamlessMode: false`)	Use `@powermem` in VS Code chat with `remember`, `save`, and `search` commands. Auto-retrieve runs before each reply; auto-summarize saves every N turns.

Toggle Seamless mode in PowerMem settings (powermem.seamlessMode).

Settings

Setting	Description	Default
`powermem.enabled`	Enable the extension	`true`
`powermem.backendUrl`	PowerMem HTTP API base URL (MCP mode: root for `{url}/mcp`)	`http://localhost:8848`
`powermem.apiKey`	API key (`X-API-Key`) if required	(empty)
`powermem.connectionMode`	`mcp` for AI tools, or `http` where a client needs HTTP context	`mcp`
`powermem.useMCP`	Deprecated; prefer `powermem.connectionMode`	`true`
`powermem.mcpServerPath`	Local MCP command/path; empty = remote MCP at `{backendUrl}/mcp`	(empty)
`powermem.userId`	User ID for memory scope; empty = auto-generated	(empty)
`powermem.projectName`	Project name; empty = workspace name	(empty)
`powermem.seamlessMode`	Zero-friction mode: linked AI + auto-capture; disable for `@powermem` chat	`true`
`powermem.autoCapture.onSave`	Add file content to memory on save	`true` in seamless mode
`powermem.autoCapture.include`	Glob patterns for auto-capture (comma-separated)	`*/.md,*/.txt,/docs/`
`powermem.autoCapture.maxChars`	Max characters per file on auto-capture	`8848`
`powermem.chat.autoSummarizeEveryNTurns`	In `@powermem` chat: auto-summarize every N turns (`0` = off)	`10`
`powermem.chat.autoRetrieve`	In `@powermem` chat: retrieve relevant memories before answering	`true`

Commands

PowerMem: Status Bar Menu — Open the main menu (link, query, add, setup, reconnect, etc.).
PowerMem: Query Memories — Search PowerMem (uses selection or prompts for query).
PowerMem: Add Selection to Memory — Save the current selection to PowerMem.
PowerMem: Quick Note — Add a one-line note to PowerMem.
PowerMem: Link to AI Tools — Write MCP/HTTP config for supported clients. Use when a broad multi-client setup is desired; otherwise configure only the current IDE/client with SETUP.md.
PowerMem: Setup — Change backend URL, API key, MCP path, test connection.
PowerMem: Dashboard — Open the simple dashboard panel.

Client Setup Details

The setup flow handles the current IDE/client first. Use the per-client guide when you need manual configuration details:

Client	Details
VS Code	`../../docs/integrations/vs_code.md`
Cursor	`../../docs/integrations/cursor.md`
Windsurf	`../../docs/integrations/windsurf.md`
GitHub Copilot	`../../docs/integrations/github_copilot.md`
Qoder	`../../docs/integrations/qoder.md`

Link to AI Tools auto-writes configs for Cursor, Claude Code, Codex, Windsurf, and GitHub Copilot when broad linking is explicitly requested. Qoder is configured through MCP; SETUP.md includes the client-specific steps. Codex and OpenCode can also use the generic MCP client flow in ../mcp-client/SETUP.md. After linking, restart or reload the respective AI tool/IDE so it picks up the new config.

Development

From source:

cd apps/vscode-extension
npm install
npm run compile

Press F5 in VS Code to launch an Extension Development Host, or package and install a .vsix:

code --install-extension powermem-vscode-*.vsix