Skip to content
Tapestry Tapestry

Connect VS Code (without Claude Code)

Copy page

VS Code (1.99+) can talk to MCP servers natively. This wires Tapestry’s two servers — your memory and the docs search — into that, so Copilot Chat and any other MCP-aware assistant in the editor can use them. No Claude Code, no extension.

Prefer one click? The Tapestry VS Code extension registers the same two servers for you. This page is the manual path — useful when you want the wiring in your repo, or you’d rather not install an extension.

Prerequisites

Section titled “Prerequisites”
  • VS Code 1.99 or newer — native MCP support.
  • For the docs server: pip install tapestry-docs-mcp, on the same python your editor can reach.
  • For the memory server: the URL of your Tapestry memory deployment (you run your own; the server is skipped cleanly if you leave it out).

Configure

Section titled “Configure”

Add the servers to a .vscode/mcp.json file. Put it in your workspace to share the wiring with a repo, or open your user-level file with the MCP: Open User Configuration command to use it everywhere.

{
  "servers": {
    "tapestry-docs": {
      "command": "python",
      "args": ["-m", "docs_mcp"]
    },
    "loom-memory": {
      "type": "http",
      "url": "https://your-memory-host.example.com/mcp/memory/"
    }
  }
}
  • tapestry-docs runs locally over stdio. If python doesn’t resolve to the environment where you ran pip install tapestry-docs-mcp, use python3 or an absolute interpreter path instead.
  • loom-memory is your own deployment. Replace the URL with yours. To use only the docs server, drop the loom-memory block entirely.

Verify

Section titled “Verify”
  1. Reload the VS Code window.
  2. Open Copilot Chat (or another MCP-aware chat surface).
  3. The Tapestry tools — tapestry_docs_search, and memory_recall / memory_write if you wired the memory server — appear in the tools list.
  4. Call tapestry_docs_search with a query like observer; it should return ranked results.

If the memory tools return 401 or 404, check that your deployment is reachable from your machine and the URL is exact.

Section titled “Related”