Harvest MCP Server
Enables AI assistants to manage Harvest time tracking by verifying credentials, listing projects and tasks, retrieving and logging time entries, and starting/stopping/restarting timers.
README
Harvest MCP Server
A Model Context Protocol (MCP) server that integrates with the Harvest time tracking API. This server enables AI assistants like Claude to interact with your Harvest account to check projects, tasks, and log time entries.
Features
- Verify credentials and retrieve authenticated user information
- List active projects with their associated tasks
- Retrieve time entries for any date or date range with optional project/user filtering
- Log time entries for specific projects and tasks
- Start/stop/restart timers to track time in real-time
- Delete time entries when needed
- Flexible authentication: Supports both environment variables and per-call credentials
Prerequisites
- Node.js (v16 or higher recommended)
- A Harvest account with API access
- Harvest Personal Access Token and Account ID
Getting Harvest Credentials
- Log in to your Harvest account
- Go to Settings → Developers
- Create a new Personal Access Token
- Note your Account ID (visible in the URL or the token creation page)
Installation
- Clone this repository:
git clone <repository-url>
cd harvest-mcp
- Install dependencies:
npm install
- Build the TypeScript code:
npm run build
Configuration
Option 1: Environment Variables (Recommended for MCP)
Configure the server in your MCP client (e.g., Claude Desktop):
Claude Desktop (~/.claude.json or %APPDATA%/Claude/config.json on Windows):
{
"mcpServers": {
"harvest": {
"command": "node",
"args": ["/absolute/path/to/harvest-mcp/build/index.js"],
"env": {
"HARVEST_ACCESS_TOKEN": "your-token-here",
"HARVEST_ACCOUNT_ID": "your-account-id"
}
}
}
}
Important:
- Use the absolute path to
build/index.js - Restart Claude Desktop after modifying the config
- Rebuild the project (
npm run build) after any code changes
Option 2: Per-Call Credentials (Stateless Mode)
If you prefer not to use environment variables, you can pass credentials with each tool invocation:
{
"access_token": "your-token-here",
"account_id": "your-account-id"
}
Available Tools
get_my_profile
Verify credentials and retrieve information about the authenticated user.
Parameters:
access_token(optional): Harvest Personal Access Tokenaccount_id(optional): Harvest Account ID
Example Response:
Authenticated as: John Doe
get_time_entries
Retrieve time entries for a specific date or date range. Supports filtering by project and user.
Parameters:
access_token(optional): Harvest Personal Access Tokenaccount_id(optional): Harvest Account IDfrom(optional): Start date in YYYY-MM-DD format (defaults to today)to(optional): End date in YYYY-MM-DD format (defaults to thefromdate)project_id(optional): Filter entries for a specific projectuser_id(optional): Filter entries for a specific user (defaults to current user)
Example Usage:
Get today's entries:
{}
Get entries for a specific date:
{
"from": "2026-02-01"
}
Get entries for a date range:
{
"from": "2026-02-01",
"to": "2026-02-05"
}
Get this week's entries for a specific project:
{
"from": "2026-02-03",
"to": "2026-02-07",
"project_id": 67890
}
Example Response:
Found 3 time entries for 2026-02-04:
Total hours: 6.50
[
{
"id": 123456789,
"spent_date": "2026-02-04",
"hours": 2.5,
"notes": "Implemented user authentication",
"project": {
"id": 67890,
"name": "My Project"
},
"task": {
"id": 222,
"name": "Development"
}
},
...
]
list_active_projects
List all active projects assigned to the authenticated user, including their associated tasks. Returns a concise, readable format by default to minimize context usage.
Parameters:
access_token(optional): Harvest Personal Access Tokenaccount_id(optional): Harvest Account IDformat(optional): Output format -'compact'(default) for concise view or'full'for complete JSON
Example Response (Compact Format - Default):
Client: Acme Corp
→ Website Redesign (project_id: 14308070)
Tasks: Design (8083367), Development (8083368), Testing (8083369)
Client: 123 Industries
→ Online Store - Phase 1 (project_id: 14308069)
Tasks: Graphic Design (8083365), Programming (8083366)
Example Response (Full Format):
[
{
"id": 12345,
"project": {
"id": 67890,
"name": "My Project",
"code": "PROJ"
},
"task_assignments": [
{
"id": 111,
"task": {
"id": 222,
"name": "Development"
}
}
],
"is_project_manager": false,
"budget": null,
...
}
]
log_time
Create a time entry for a specific project and task.
Parameters:
access_token(optional): Harvest Personal Access Tokenaccount_id(optional): Harvest Account IDproject_id(required): Project IDtask_id(required): Task IDhours(required): Number of hours to logspent_date(optional): Date in YYYY-MM-DD format (defaults to today)notes(optional): Description of the work performed
Example Usage:
{
"project_id": 67890,
"task_id": 222,
"hours": 2.5,
"spent_date": "2026-02-04",
"notes": "Implemented user authentication"
}
Example Response:
Success! Time entry created with ID: 123456789
start_timer
Start a running timer for a specific project and task. The timer will track time until stopped.
Parameters:
access_token(optional): Harvest Personal Access Tokenaccount_id(optional): Harvest Account IDproject_id(required): Project IDtask_id(required): Task IDspent_date(optional): Date in YYYY-MM-DD format or relative date (defaults to today)notes(optional): Description of the work being performed
Example Usage:
{
"project_id": 67890,
"task_id": 222,
"notes": "Working on user authentication"
}
Example Response:
Success! Timer started with ID: 123456789
stop_timer
Stop a running timer. This uses Harvest's dedicated /stop endpoint to calculate hours based on elapsed time.
Parameters:
access_token(optional): Harvest Personal Access Tokenaccount_id(optional): Harvest Account IDtime_entry_id(required): The ID of the running timer to stop
Example Usage:
{
"time_entry_id": 123456789
}
Example Response:
Success! Timer stopped.
Project: My Project
Task: Development
Total time logged: 2.5 hours
Date: 2026-02-12
restart_timer
Restart a previously stopped timer. This uses Harvest's dedicated /restart endpoint to resume tracking.
Parameters:
access_token(optional): Harvest Personal Access Tokenaccount_id(optional): Harvest Account IDtime_entry_id(required): The ID of the stopped timer to restart
Example Usage:
{
"time_entry_id": 123456789
}
Example Response:
Success! Timer restarted.
ID: 123456789
Project: My Project
Task: Development
Date: 2026-02-12
get_running_timer
Check if there's a timer currently running and see elapsed time.
Parameters:
access_token(optional): Harvest Personal Access Tokenaccount_id(optional): Harvest Account ID
Example Response:
Timer is running:
ID: 123456789
Project: My Project
Task: Development
Notes: Working on user authentication
Elapsed: 1h 23m
[detailed JSON data]
delete_time_entry
Delete a specific time entry.
Parameters:
access_token(optional): Harvest Personal Access Tokenaccount_id(optional): Harvest Account IDtime_entry_id(required): The ID of the time entry to delete
Example Usage:
{
"time_entry_id": 123456789
}
Example Response:
Success! Time entry 123456789 has been deleted.
Development
Build the project
npm run build
Run in development mode (with ts-node)
npm run dev
Type checking only
npx tsc --noEmit
Project Structure
harvest-mcp/
├── index.ts # Main server implementation
├── build/ # Compiled JavaScript output
│ └── index.js # Entry point for MCP clients
├── package.json # Dependencies and scripts
├── tsconfig.json # TypeScript configuration
└── README.md # This file
Usage with Claude Desktop
Once configured, you can use natural language to interact with Harvest:
Example interactions:
- "What projects am I assigned to in Harvest?"
- "Show me my time entries for today"
- "How many hours have I logged today?"
- "Show me my time entries for last Monday" (2026-02-03)
- "Get all time entries from February 1st to February 5th"
- "Show me time entries for project 67890 this week"
- "Log 3 hours to project 12345, task 222 with notes 'Fixed authentication bug'"
- "Start a timer for project 67890, task 222"
- "Is a timer running?"
- "Stop my current timer"
- "Restart timer 123456789"
- "Delete time entry 123456789"
- "Show me my Harvest profile"
Troubleshooting
Server not connecting
- Ensure the build step completed successfully (
npm run build) - Verify the absolute path in your MCP config points to
build/index.js - Restart your MCP client after configuration changes
Authentication errors
- Double-check your Personal Access Token and Account ID
- Ensure credentials are correctly set in environment variables or passed with tool calls
- Verify your Harvest account has API access enabled
Missing projects or tasks
- Ensure you're assigned to the projects in Harvest
- Only active projects are returned by
list_active_projects - Check that your user has the necessary permissions
API Reference
This server uses the Harvest API v2. Key endpoints:
GET /users/me- Current user informationGET /users/me/project_assignments- User's project assignmentsGET /time_entries- Retrieve time entriesPOST /time_entries- Create a time entry or start a timerPATCH /time_entries/{id}/stop- Stop a running timerPATCH /time_entries/{id}/restart- Restart a stopped timerDELETE /time_entries/{id}- Delete a time entry
License
ISC
Author
Sebastiaan Pfennigwerth
Contributing
Contributions are welcome! Please feel free to submit issues or pull requests.
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。