Humanforai MCP server
{
"mcpServers": {
"humanforai": {
"command": "npx",
"args": [
"-y",
"humanforai@latest"
]
}
}
}{
"mcpServers": {
"humanforai": {
"command": "npx",
"args": [
"-y",
"humanforai@latest"
]
}
}
}{
"mcpServers": {
"humanforai": {
"command": "npx",
"args": [
"-y",
"humanforai@latest"
]
}
}
}{
"servers": {
"humanforai": {
"command": "npx",
"args": [
"-y",
"humanforai@latest"
]
}
}
}{
"mcpServers": {
"humanforai": {
"command": "npx",
"args": [
"-y",
"humanforai@latest"
]
}
}
}file .mcp.json (in your project root)file claude_desktop_config.jsonfile ~/.cursor/mcp.jsonfile .vscode/mcp.jsonfile ~/.codeium/windsurf/mcp_config.jsoninstall npx -y humanforaiversion 0.1.4
What it does
This server connects AI agents to real human operators for tasks requiring physical presence, human perception, or judgment. It supports real-world verification, product testing, AI output review, and local errands, with responses typically within 12 hours.
- Verify that a product, place, or price claim exists by having a human visit, photograph, and report back with evidence.
- Have a human test an app or product by installing it, using it, and providing detailed feedback on functionality and user experience.
- Get human review of AI-generated content or decisions before they reach production to assess tone, clarity, trustworthiness, or design quality.
Tools this server exposes
| Tool | Does |
|---|---|
| get_human_services | Retrieve the catalog of available services, accepted task types, and operational limits. |
| submit_human_task | Submit a task to a human operator and receive a task_id for tracking. |
| check_task_status | Poll the status of a submitted task by task_id, including when the operator saw it, estimated delivery time, and status history. |
| message_human_operator | Ask questions or discuss task scope with the human operator before formal submission. |
Before you enable it
An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.