# Tendem MCP server [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://github.com/Toloka/tendem-mcp  
**GitHub Stars:** 101  
**Views:** 1  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/tendem-mcp-server

## Description
Delegate tasks to vetted human experts - research, writing, analysis, and data work.

## Claude Desktop Quick Installation
Remote MCP endpoint (confidence: high). Install path detected from listing signals. Add as a URL/SSE server in your client:

```json
"mcpServers": {
  "tendem-mcp-server": {
    "url": "https://mcp.tendem.ai/mcp"
  }
}
```

## Documentation

## What Tendem MCP server MCP server does

Tendem MCP server MCP server lets an AI assistant send work to Tendem’s network of vetted human experts. The service is intended for tasks where automated reasoning may need outside judgment or hands-on execution, including research, competitive analysis, copywriting, editing, design review, data cleaning, and complex multi-step assignments.

The repository distributes client plugins, skills, rules, and connection manifests. It does not contain the MCP service itself. The hosted endpoint is `https://mcp.tendem.ai/mcp`, and task outputs can include reports, markdown, structured data, and downloadable files. The Tendem web application remains available for account management, balance top-ups, and viewing completed task output.

## How it works

A typical workflow starts when an agent creates a task in plain language. Tendem then scopes the request and provides a quote. The agent can answer follow-up questions, review the proposed scope, and approve the task. Approval triggers the charge, after which the agent polls for progress and retrieves the completed result.

The Tendem MCP server MCP server exposes eleven tools for this lifecycle: `create_task`, `get_task`, `get_contract`, `get_task_result`, `list_tasks`, `approve_task`, `cancel_task`, `send_message`, `read_chat`, `get_account`, and `get_file_upload_url`. Plugin guidance adds lifecycle instructions, slow polling behavior, scope negotiation, and file transfer guidance. Claude Code and GitHub Copilot can also use a background watcher and desktop notifications when a task needs input, has a quote, requires a top-up, or is ready.

Automated workflows should check `get_account` before approval and place their own approval gate around `approve_task`. If the balance is insufficient, the service returns a task-specific top-up URL rather than an error intended for retry.

## Setup and configuration

Interactive clients connect to the remote server over streamable HTTP and complete OAuth on first use. A generic MCP configuration uses the `http` type and the hosted URL. The repository provides installation paths for Claude Code, OpenAI Codex or ChatGPT, GitHub Copilot CLI, Gemini CLI, and Kiro. It also includes client-specific plugin, extension, and Power files.

For headless or programmatic use, create an API key from the Agent builders area at `agent.tendem.ai/mcp`. Python, LangChain, and LangGraph users can use the included `langchain-tendem` integration, which provides four price-capped tools and handles polling in Python. Other stacks can use `mcp-remote` as a local stdio bridge to pass an `Authorization: ApiKey` header to the hosted endpoint. Keep that key in a secret store or injected environment variable rather than source control.

## Limitations and notes

The Tendem MCP server MCP server is remote; this repository does not provide a locally hosted implementation. Interactive OAuth is not suitable for every CI or headless setup, so those environments need API-key authentication and an appropriate bridge or integration. Charges occur when a task is approved, and human work may require additional communication or scope decisions before a result is available. The repository states that the full task output is also available through the Tendem web interface.

_Full upstream README: https://allmcps.com/mcp/tendem-mcp-server/readme_