mcpfold

MCP config sprawls across clients (Claude Code, Cursor, VS Code, Windsurf, Zed, …), and the formats have quietly diverged: VS Code uses the root keyservers, Zed usescontext_servers, everyone else usesmcpServers. Secrets get hardcoded into plaintext JSON.mcpfoldkeeps one canonical file and folds it out to each client — resolving secretreferences(never values) and curating which servers and tools each client loads.

npm / npx— no install needed to try it:

npx mcpfold init npm install -g mcpfold # or: pnpm add -g mcpfold (installs mcpfold + the mcpf alias)

scoop bucket add mcpfold https://github.com/dj-pearson/scoop-bucket scoop install mcpfold

curl | sh(macOS / Linux) — standalone binary, no Node, checksum-verified:

curl -fsSL https://mcpfold.com/install.sh | sh

RequiresNode 20+(for the npm/npx install; the binaries need nothing).mcpfoldauto-detects whichever MCP clients you have installed.

mcpfold init # 1. scaffold a commented mcp.config.jsonc (+ $schema for editor autocomplete) mcpfold import # 2. scan installed clients and merge their servers into the canonical file mcpfold sync # 3. fold the canonical config out to every detected client (native formats) mcpfold diff # preview what sync would change, per client, before applying mcpfold doctor # health-check config, clients, and secret references

The savings above come fromcuration— exposing only the tools an agent actually uses.curatewrites an allow-list;syncfolds each curated server out as amcpfold runproxy shim, so clients only ever load the trimmed toolset:

mcpfold add fs --package @modelcontextprotocol/server-filesystem # add a server mcpfold curate fs --tools read_text_file,write_file,list_directory # keep only what you use # → fs: keeping 3 of 14 tools, ~3.2k → ~684 tokens (approx) mcpfold sync # fold it out (writes a proxy shim)

Already been running your servers? Plainmcpfold curatereads the local audit trail and recommends the allow-list from youractualtool usage. Check the measured savings on your own config anytime withmcpfold status.

18 clients, each read and written in its own native format from a single source of truth:

Claude Code · Claude Desktop · Cursor · VS Code · Visual Studio · Windsurf · Zed · Cline · Continue · Roo Code · Gemini CLI · Codex CLI · Copilot CLI · JetBrains · Goose · LM Studio · Warp · opencode

mcpfold initadds a$schemaline so editors give you autocomplete + inline validation:

{ "$schema": "https://mcpfold.com/schema/v1.json", "version": 1, // … }

Core purityis enforced:packages/coremay not importnode:fs,node:os,node:path, or any network/process library. All I/O is injected throughClientAdapter/SecretProvider. Guarded by an ESLintno-restricted-importsrule and thescripts/check-core-purity.mjsCI gate.

RequiresNode 20+andpnpm 10+(corepack enable).

pnpm install # install workspace deps pnpm lint # eslint + core-purity check pnpm typecheck # tsc --noEmit across packages pnpm test # vitest (unit + fixture snapshots) pnpm -r build # build every package pnpm verify_all # lint + typecheck + test + build (the full gate)

If mcpfold saves you time, you can help fund ongoing work:

Sponsorships fund the free, open-source core. Thank you 🙏

This is a web browser that enables your coding agent, such as Claude Code, to visit websites on your behalf and assist you in identifying bugs or creating UI test cases.

Create crafted UI components inspired by the best 21st.dev design engineers.

Bring agent evaluations, observability, and synthetic test set generation directly into your IDE for free with Galileo's new MCP server

An MCP server to help AI assistants to answer questions and generate AccelByte Extend SDK code more effectively .

MCP server for AI Diagram Maker — generate beautiful software engineering diagrams directly inside Cursor, Claude Desktop, Claude Code, or any MCP-compatible AI agent

ALAPI MCP Tools,Call hundreds of API interfaces via MCP

AI-powered SVG animation generator that transforms static files into animated SVG components using the Allyson platform

MCP server that gives AI assistants on-demand access to 1,500+ amCharts docs, ~300 code examples, and 1000+ class API references.

APIMatic MCP Server is used to validate OpenAPI specifications using APIMatic. The server processes OpenAPI files and returns validation summaries by leveraging APIMatic’s API.

One shared context layer for AI agents and humans — live API specs, DB schemas, and versioned contracts across repos so every agent and teammate works from the same source of truth.