Dev Tool Config Sync MCP Server
Syncs and manages configurations for dev tools like Claude Code and Gemini CLI across workspaces with snapshot history and rollback capabilities.
README
Dev Tool Config Sync MCP Server
An MCP server that syncs dev tool configurations to/from a storage backend with snapshot history and rollback. Tool-agnostic — declarative YAML profiles define what to sync for each tool (Claude Code, Gemini CLI, opencode, etc.).
Designed to run as a sidecar container in Eclipse Che workspaces, but works standalone.
Quick start
npm install
npm run build
npm start
The server listens on port 8089 (configurable via CONFIG_SYNC_PORT) and exposes MCP tools over streamable HTTP at /mcp. Health endpoint at /healthz.
Environment variables
| Variable | Default | Description |
|---|---|---|
CONFIG_SYNC_PORT |
8089 |
HTTP port |
CONFIG_SYNC_STORAGE_DIR |
~/.config-sync-storage |
File-based storage directory |
CONFIG_SYNC_PROFILES_DIR |
./profiles |
Directory containing tool profile YAML files |
CONFIG_SYNC_USER_ID |
CHE_USER_ID or default |
User identifier for storage key namespacing |
FILE_ACCESS_MODE |
local |
File access mode: local (direct filesystem) or remote (exec into workspace pods) |
POD_NAMESPACE |
auto-detected | Kubernetes namespace for pod discovery (auto-detected from /var/run/secrets/kubernetes.io/serviceaccount/namespace) |
TARGET_WORKSPACE |
none | Optional default workspace name for remote mode |
WORKSPACE_HOME_DIR |
~ |
Home directory path inside workspace pods |
MCP tools
| Tool | Description |
|---|---|
sync_from_storage |
Pull latest config from storage backend (all tools or a specific tool) |
sync_to_storage |
Push current workspace config to storage backend (all tools or a specific tool) |
list_config_versions |
List available config snapshots/versions for a tool |
rollback_config |
Restore config from a previous version for a tool |
diff_config |
Show differences between local and stored config for a tool |
get_sync_status |
Check last sync time and pending changes (all tools or a specific tool) |
list_tools |
List all registered tool profiles and their sync status |
Tool profiles
Each supported tool has a YAML profile in profiles/ defining its sync rules:
tool: claude-code
name: Claude Code
paths:
sync:
- "~/.claude/settings.json"
- "~/.claude/settings.local.json"
- "~/.claude/CLAUDE.md"
- "~/.claude/agents/**"
- "~/.claude/hooks/**"
- "~/.claude/plugins/**"
- "~/.claude.json"
skip:
- "~/.claude/plugins/cache/**"
sensitive:
- "**/*credentials*"
- "**/*token*"
- "**/*.key"
Built-in profiles: claude-code, gemini-cli, opencode. Add new tools by dropping a YAML file in the profiles directory — no code changes needed.
Profile schema
| Field | Required | Description |
|---|---|---|
tool |
yes | Unique identifier (^[a-z][a-z0-9-]*$), used as storage key |
name |
yes | Human-readable display name |
paths.sync |
yes (≥1) | Glob patterns relative to $HOME to sync |
paths.skip |
no | Glob patterns to exclude from sync |
paths.sensitive |
no | Glob patterns for files that must never be synced (credentials, tokens) |
All paths must start with ~/. Symlinks pointing outside $HOME are rejected. Max file size: 1 MB (configurable).
Architecture
┌──────────────────────────────────────────────────────┐
│ Workspace Pod │
│ │
│ ┌──────────────┐ │
│ │ Claude Code │─MCP─┐ │
│ └──────────────┘ │ ┌───────────────────────┐ │
│ ┌──────────────┐ ├──│ Config Sync Server │ │
│ │ Gemini CLI │─MCP─┤ │ (sidecar, port 8089) │ │
│ └──────────────┘ │ │ │ │
│ ┌──────────────┐ │ │ profiles/ │ │
│ │ other tools │─MCP─┘ │ claude-code.yaml │ │
│ └──────────────┘ │ gemini-cli.yaml │ │
│ │ opencode.yaml │ │
│ └──────────┬────────────┘ │
│ │ │
└──────────────────────────────────────┼───────────────┘
│
┌─────────▼─────────┐
│ Storage Backend │
│ (file-based) │
└───────────────────┘
Storage
Each push creates an immutable snapshot at {storageDir}/{userId}/{tool}/{version}/, containing a manifest.json (checksums, sizes, permissions) and the raw files. Snapshots are append-only. Storage key uses userId + tool — configs follow the user across workspaces, stored within the user namespace on the cluster.
Conflict strategy: Last-write-wins. Use diff_config to inspect drift before overwriting.
What gets synced vs skipped
- Sync: user-edited configuration, preferences, custom extensions (agents, hooks, plugins)
- Skip: session state, history/transcripts, caches, analytics, anything regeneratable or workspace-local
Docker
npm run build
docker build -t config-sync-mcp .
docker run -p 8089:8089 config-sync-mcp
Base image: registry.access.redhat.com/ubi10/nodejs-24-minimal
Cluster deployment
Deploy the centralized config-sync-mcp server to a Kubernetes/OpenShift namespace where DevWorkspaces run.
Prerequisites
ocorkubectlconfigured for the target cluster- Namespace where DevWorkspaces run (e.g.,
kubeadmin-devspaces)
Apply manifests
oc project <namespace>
oc apply -f deploy/
This creates:
- ServiceAccount
config-sync-mcp— identity for the server pod - Role — permissions to list pods, exec into workspace pods, and watch DevWorkspaces
- RoleBinding — binds the role to the service account
- Deployment — the server running in remote mode (
FILE_ACCESS_MODE=remote) - Service — exposes the MCP endpoint at
config-sync-mcp:8089
Verify
# Check pod is running
oc get pods -l app=config-sync-mcp
# Check health
oc exec deployment/config-sync-mcp -- curl -s http://localhost:8089/healthz
# Test MCP handshake
oc exec deployment/config-sync-mcp -- curl -s -X POST http://localhost:8089/mcp \
-H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-d '{"jsonrpc":"2.0","method":"initialize","params":{"protocolVersion":"2025-03-26","capabilities":{},"clientInfo":{"name":"test","version":"0.1"}},"id":1}'
Development
npm install
npm run build # compile TypeScript + copy profiles to dist/
npm test # run all tests (vitest)
npm run test:watch # watch mode
Project structure
src/
index.ts entrypoint
server.ts HTTP server + MCP transport
types.ts shared types and interfaces
tools/ MCP tool handlers (one per tool)
storage/ storage backend interface + file implementation
profiles/ YAML profile loader + glob resolver
sync/ sync engine + SHA-256 checksums
profiles/ built-in tool profile YAML files
tests/ unit + integration tests
Next steps
- [ ] Auto-sync on workspace start/stop via lifecycle hooks
- [ ] Test concurrent workspace scenarios
- [ ] Admin-pushed baseline profiles
- [ ] Add opencode built-in profile
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。