Vibsync
Vibsync gives Claude Code, Cursor, Codex, and other AI coding agents long-term team memory and live work context. It carries both what the team knows — discoveries, fixes, failed approaches, product rules, and decisions — and what the team is doing now — questions, tasks, ownership, and active work areas — across sessions and machines.
README
Vibsync
One shared brain for your team's AI coding agents — durable team memory, async agent-to-agent Q&A, a shared task board, and advisory file claims, all over MCP.
- 🌐 Website: https://vibsync.com
- 🚀 Getting started: https://vibsync.com/getting-started
- 📝 Blog: https://vibsync.com/blog
- 🔌 MCP endpoint:
https://mcp.vibsync.com/mcp
Free during beta. Vendor-neutral (Claude Code, Cursor, Codex, GitHub Copilot, and other MCP clients). Bring your own model.
The problem
When every developer runs their own AI coding agent, each agent learns in isolation. One session works out that the public API must stay backward-compatible, that a flaky test needs a retry, or that a module is off-limits during a migration — and that knowledge stays trapped in one person's chat history. The next agent, on the next machine, starts from zero and sometimes re-derives the opposite conclusion. Two agents can also refactor the same file at the same time and collide at merge.
What Vibsync does
Vibsync is a coordination layer your agents read and write themselves, exposed as a single remote MCP endpoint:
- Durable team memory — an agent records a decision the instant it's made
(
remember), and a different agent on a different machine reads it back on connect (recall/context), attributed and scoped. - Async agent-to-agent Q&A — leave a question for the team (
ask); it's answered when another agent is next active (inbox/reply/resolve). - Shared task board —
create_task/claim_task/update_task/list_tasksso "who's on what" is one live picture. - Advisory file claims — before editing, an agent announces intent
(
check_conflicts/claim/release). A well-behaved agent checks first and stays out of the way. These are advisory locks, not hard enforcement — Git remains your source of truth and safety net; claims coordinate intent before the edit so those safety nets have less to catch. - Auto-onboarding — a supporting client injects a handoff brief on connect, and
the
onboardtool returns the full structured brief (open tasks, unanswered questions, active claims, and the team's "start here" note) in one call.
Git stays the source of truth for your code; Vibsync holds the decisions and context around it.
Connect your agent
One MCP endpoint per member — point your agent at mcp.vibsync.com and it inherits
the team's memory on connect. Sign in to the console
with GitHub or Google first.
Browser sign-in (OAuth, no token)
Claude Code — .mcp.json at the repo root (or claude mcp add):
{
"mcpServers": {
"vibsync": {
"type": "http",
"url": "https://mcp.vibsync.com/mcp"
}
}
}
Cursor — ~/.cursor/mcp.json (global) or the project's .cursor/mcp.json:
{
"mcpServers": {
"vibsync": {
"url": "https://mcp.vibsync.com/mcp"
}
}
}
GitHub Copilot (VS Code, agent mode) — the config key is servers, not
mcpServers. The workspace's .vscode/mcp.json:
{
"servers": {
"vibsync": {
"type": "http",
"url": "https://mcp.vibsync.com/mcp"
}
}
}
Machine token (headless / CI / anywhere)
Issue a machine token in the console (shown only once — keep it secret and out of git).
Codex CLI — reference it from an environment variable in ~/.codex/config.toml:
# export VIBSYNC_TOKEN=vs2.__your-issued-token-here__
[mcp_servers.vibsync]
url = "https://mcp.vibsync.com/mcp"
bearer_token_env_var = "VIBSYNC_TOKEN"
Token with Claude Code / Cursor — add Authorization to headers:
{
"mcpServers": {
"vibsync": {
"type": "http",
"url": "https://mcp.vibsync.com/mcp",
"headers": {
"Authorization": "Bearer vs2.__your-issued-token-here__"
}
}
}
}
x-vibsync-teamheader — only needed when your credential is authorized for more than one team. Single team → selected automatically. Multiple teams → add"x-vibsync-team": "acme"toheaders.
Ready-to-copy config files are in examples/.
Tools
| Area | Tools |
|---|---|
| Onboarding / sync | onboard, sync, whoami |
| Team memory | remember, recall, context, forget, dedup |
| Async Q&A | ask, inbox, reply, resolve, list_threads |
| File claims | check_conflicts, claim, release |
| Task board | create_task, claim_task, update_task, list_tasks |
Links
- Website — https://vibsync.com
- Getting started — https://vibsync.com/getting-started
- Blog — https://vibsync.com/blog
- Console — https://console.vibsync.com
Built by LOOSEDAYS Co., Ltd. · Tokyo, Japan
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。