VS Code Agent Mode supports MCP servers through GitHub Copilot. Engram works as a remote HTTP MCP server and does not require a local stdio bridge for VS Code.
curl -fsSL https://engram-memory.com/install | sh- Create or edit your VS Code user-profile
mcp.json:- macOS:
~/Library/Application Support/Code/User/mcp.json - Linux:
~/.config/Code/User/mcp.json - Windows:
%APPDATA%\Code\User\mcp.json
- macOS:
{
"servers": {
"engram": {
"type": "http",
"url": "https://www.engram-memory.com/mcp"
}
}
}VS Code also supports workspace MCP config at .vscode/mcp.json. Use the user-profile
file for Engram if you want it available across all workspaces.
For local development:
{
"servers": {
"engram": {
"type": "http",
"url": "http://localhost:7474/mcp"
}
}
}- Restart VS Code
- Ensure GitHub Copilot extension is installed
- If VS Code prompts you to trust the MCP server, approve Engram after reviewing the server URL.
- Open Chat in Agent Mode.
- Tell it:
"Set up Engram for my team"to create a workspace - Or:
"Join Engram with key ek_live_..."to join existing workspace
With Copilot + Engram, your AI assistant will:
- Query team knowledge before starting tasks
- Commit important discoveries
- Detect conflicts in team knowledge
engram verifyIn your IDE: Ask your agent: "Call engram_status and tell me what it returns."
Expected output:
{"status": "ready", "mode": "team", "engram_id": "ENG-XXXXXX", "schema": "engram"}
- Ensure GitHub Copilot is active
- Confirm Agent Mode is enabled in Chat
- Check MCP config path matches your OS
- Confirm the Engram entry is under
servers.engram - Restart VS Code after config changes
- Use VS Code's MCP output log if the server appears but does not start
See docs/TROUBLESHOOTING.md for more help.