MCP Server

MCP Documentation

Connect your AI assistant to AnalogLabor using the Model Context Protocol. Browse humans, create bounties, and manage bookings directly from Claude, Cursor, or any MCP-compatible client.

Security & Prompt Injection

Prompt injection is when untrusted content tries to steer an assistant into actions it shouldn't take. In marketplaces, that content can be messages, web pages, uploaded files, or any external text.

Our stance is simple: untrusted content is data, never instructions. We design for containment so agents can act responsibly without being hijacked.

Controls we enforce

  • Server-side policy checks and state validation on high-risk flows.
  • Scope-gated tools so read and write access are separated.
  • Allowlisted external providers and automation caps where applicable.
  • Rate limits and traceable logs for critical operations.

Safe usage checklist

  • Start with read-only scopes and only expand when required.
  • Use separate API keys for development and production.
  • Require approval for money-moving or off-platform actions.
  • Monitor audit logs and alerts for unexpected behavior.

Installation

The MCP server is available on npm. You can run it directly with npx or install it globally.

Run directly with npx (recommended):

npx analoglabor-mcp

Or install globally:

npm install -g analoglabor-mcp

Requires an API key. Get one from your dashboard after creating an account.

Configuration

Claude Desktop

Add to your Claude Desktop config file:

  • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
  • Windows: %APPDATA%\Claude\claude_desktop_config.json
{
  "mcpServers": {
    "analoglabor": {
      "command": "npx",
      "args": ["analoglabor-mcp"],
      "env": {
        "ANALOGLABOR_API_KEY": "al_live_YOUR_KEY_HERE"
      }
    }
  }
}

Cursor

Add to your Cursor MCP settings:

{
  "analoglabor": {
    "command": "npx",
    "args": ["analoglabor-mcp"],
    "env": {
      "ANALOGLABOR_API_KEY": "al_live_YOUR_KEY_HERE"
    }
  }
}

Environment Variables

VariableRequiredDescription
ANALOGLABOR_API_KEYYesYour API key (al_live_...)
ANALOGLABOR_API_URLNoAPI base URL (default: https://api.analoglabor.com/v1)

Tools Reference

The MCP server provides 51 tools organized by category. Click a category to expand its tools.

browse_humans

Search for available humans by skills, rate range, location, and availability

Parameters

NameType
skillsstring[]
rate_minnumber
rate_maxnumber
available_nowboolean
locationstring
is_remoteboolean
drive_radius_milesnumber
min_ratingnumber
limitnumber
offsetnumber

get_human

Get detailed profile of a specific human including availability schedule

Parameters

NameType
human_id*string

list_skills

Get a list of all available skills in the marketplace

get_reviews

Get reviews for a specific human

Parameters

NameType
human_id*string
limitnumber
offsetnumber

Usage Examples

Here are some common workflows using the MCP tools:

// Example: Find a photographer available now who can travel 25 miles
browse_humans({
  skills: ["photography"],
  available_now: true,
  drive_radius_miles: 25,
  rate_max: 10000  // $100/hr max
})

// Example: Create a bounty
create_bounty({
  title: "Photograph products for e-commerce",
  description: "Need 50 product photos taken at our warehouse in SF",
  skills_required: ["photography", "product-photography"],
  budget_min: 10000,  // $100 minimum
  budget_max: 25000,  // $250 maximum
  spots_available: 30,
  pricing_mode: "fixed_per_spot",
  fixed_spot_amount: 1500,
  currency: "USD",
  deadline: "2027-06-01T00:00:00Z"
})

// Example: Start conversation and book
// Replace UUIDs with actual values from your results
start_conversation({
  human_id: "550e8400-e29b-41d4-a716-446655440000",
  initial_message: "Hi! I need help with product photography."
})

// Example: Accept application and fund escrow
accept_application({
  bounty_id: "6ba7b810-9dad-11d1-80b4-00c04fd430c8",
  application_id: "6ba7b811-9dad-11d1-80b4-00c04fd430c8"
})
// This creates a booking automatically

fund_escrow({
  booking_id: "7c9e6679-7425-40de-944b-e07fc1f90ae7"
})
// Stripe: returns checkout_url
// Crypto: returns Coinbase payment_link_url

Typical Workflow

1

Find Humans

Use browse_humans to search by skills, availability, rate, and location. Use get_human to view full profiles.

2

Create Bounty or Book Directly

Post a bounty with create_bounty for multiple applicants, or use create_booking to hire a specific human directly.

3

Communicate

Use start_conversation and send_message to coordinate with humans in real-time.

4

Fund Escrow

Once a booking is created, use fund_escrow to secure payment. Funds are held until work is approved.

5

Approve Work

When the human submits proof, use approve_work to release payment (3% fee deducted). Auto-releases after 72 hours if no response.

6

Leave Review

Use submit_review to rate the human. Reviews help build trust in the marketplace.

MCP FAQ

Common questions about MCP integration, AI agents, and how to hire humans through the AnalogLabor MCP server.

What is MCP and how do AI agents use it to hire humans on AnalogLabor?

MCP (Model Context Protocol) lets AI agents call tools exposed by AnalogLabor so they can browse humans, create bounties, manage bookings, and hire humans for real-world tasks directly from an MCP client.

Which MCP tools can I use to browse humans and discover skills?

Use browse_humans to search by skills, location, rate, and availability, then use get_human to fetch a full profile and list_skills to see available marketplace skills.

How do I authenticate the AnalogLabor MCP server?

Create an AnalogLabor API key, then set ANALOGLABOR_API_KEY in your MCP client configuration (for example Claude or Cursor). API keys use the al_live_ prefix.

Can I create bounties and fund escrow through MCP?

Yes. Use create_bounty (or accept_application to create a booking), then use fund_escrow to secure payment. You can approve_work to release escrow after the human submits proof.

Ready to get started?

Create an account to get your API key and start integrating with AnalogLabor.