temporal-mcp-server
Enables AI assistants to observe and manage Temporal workflows, including listing, starting, signaling, resetting, terminating, querying, and retrieving history, with support for wait_for_activity and webhook utilities.
README
Temporal MCP Server
A Model Context Protocol (MCP) server that lets AI assistants observe and manage Temporal workflows. Built in Kotlin/JVM with a Node.js bridge for fast MCP initialization.
Prerequisites
- JDK 17+
- Node.js (for the MCP bridge)
- Access to a Temporal server (local or Temporal Cloud)
Build
./gradlew shadowJar
The JAR is created at build/libs/temporal-mcp-server-1.0-SNAPSHOT.jar.
Configuration
The server is launched via minimal-mcp.js, which handles MCP protocol initialization and spawns the Java process.
MCP Client Configuration
{
"mcpServers": {
"temporal": {
"command": "node",
"args": [
"path/to/minimal-mcp.js",
"--namespace", "my-namespace",
"--address", "my-host:7233"
],
"env": {
"TEMPORAL_API_KEY": "your-api-key"
}
}
}
}
Command-Line Arguments
| Argument | Required | Description |
|---|---|---|
--namespace |
Yes | Temporal namespace |
--address |
Yes | Temporal server address (e.g., my-ns.abc.tmprl.cloud:7233) |
Environment Variables
| Variable | Description |
|---|---|
TEMPORAL_API_KEY |
API key for Temporal Cloud authentication |
TEMPORAL_CLIENT_CERT_PATH |
Path to mTLS client certificate file |
TEMPORAL_CLIENT_KEY_PATH |
Path to mTLS client key file |
TEMPORAL_CLIENT_CERT |
mTLS client certificate content (alternative to path) |
TEMPORAL_CLIENT_KEY |
mTLS client key content (alternative to path) |
Authentication
API Key (Temporal Cloud)
Set TEMPORAL_API_KEY in the env section of your MCP config. The key is sent as a Bearer token in gRPC metadata.
mTLS Certificates
For mTLS authentication, provide either file paths or content directly:
{
"mcpServers": {
"temporal": {
"command": "node",
"args": ["path/to/minimal-mcp.js", "--namespace", "my-ns", "--address", "my-host:7233"],
"env": {
"TEMPORAL_CLIENT_CERT_PATH": "/path/to/client.pem",
"TEMPORAL_CLIENT_KEY_PATH": "/path/to/client.key"
}
}
}
}
Tools
Workflow Observation
| Tool | Description |
|---|---|
list_workflows |
Lists workflows in the namespace. Supports query filter and pageSize (1-1000, default 10). |
get_workflow_status |
Gets status of a specific workflow by workflowId (optional runId). Returns parentWorkflowId, parentRunId, taskQueue, closeTime, memo, and searchAttributes. |
get_workflow_history |
Retrieves the complete event history of a workflow with full event attributes. |
query_workflow |
Executes a query on a running workflow (workflowId, queryType, optional args). |
Workflow Actions
| Tool | Description |
|---|---|
start_workflow |
Starts a new workflow (workflowId, workflowType, taskQueue, optional args). |
send_signal |
Sends a signal to a running workflow (workflowId, signalName, optional signalArgs). |
reset_workflow |
Resets a workflow to a specific point in its history, creating a new run. Supports eventId or resetType: "last_workflow_task" to auto-detect the last WorkflowTaskCompleted event. |
terminate_workflow |
Terminates a running workflow (workflowId, optional reason). |
Utilities
| Tool | Description |
|---|---|
wait_for_activity |
Polls until a workflow reaches a specific activity (workflowId, activityName, optional timeoutSeconds 1-3600). |
send_webhook |
Sends an HTTP POST webhook (webhookUrl, payload, optional headers). Useful for completing async activities. |
Development
Running Tests
./gradlew test
Architecture
minimal-mcp.js acts as a bridge: it responds immediately to the MCP initialize handshake, then spawns the Java process in the background. Once Java is ready, all subsequent requests are forwarded to it.
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。