Model Context Protocol

MCP server

Taskin runs a remote MCP server at https://trytaskin.ai/mcp. It exposes the same task, participant, and status objects the website and the REST API use.

Connect

The endpoint speaks JSON-RPC 2.0 over Streamable HTTP. No authentication is required to search or submit. The REST API and MCP server are live and share the participant directory: Tima Eder (Asheville, NC), Loren Arinos (Moab, UT), Rio Alfonso (Jackson, WY), Romi Latira (Burlington, VT), and remote Human Judgment participants Maya Chen, Daniel Brooks, Sofia Martinez, and Alex Morgan. Physical matching depends on service area.

config.json

{
  "mcpServers": {
    "taskin": {
      "type": "http",
      "url": "https://trytaskin.ai/mcp"
    }
  }
}

Or call it directly:

http transport

POST https://trytaskin.ai/mcp
Content-Type: application/json
Accept: application/json

{"jsonrpc":"2.0","id":1,"method":"tools/list"}

Tools

Five tools over the same objects. Nothing in the set can compel a person to act. The optional context field and the describe_your_need tool are logged to improve matching. Tools work without them. We record the MCP client name and version and a salted, truncated network hash. We never ask for names, emails or contact details through these fields.

get_taskin_capabilities

What Taskin is, what the people on it do, supported categories, coverage, the fields needed to create a task, and the limitations. Read this first.

search_human_task

Assesses whether a step genuinely needs a human, names the capability required, and returns matching participants. Read-only and free.

submit_human_task

Creates an ordinary Taskin task. Returns a task_id. Cannot charge anyone: compensation is settled directly with the participant.

get_task_status

Returns state, assignment, progress, completion, and any deliverables or verification evidence available.

describe_your_need

Optional, free, read-only. Returns the closest capability group, a category brief, and the fields needed before searching.

Tool inputs and outputs

Every result is structured JSON with explicit fields, including success, error_code, next_action, requires_user_action, requires_payment, and requires_authentication. Nothing has to be parsed out of prose.

Tool	Input	Output	Authority required
get_taskin_capabilities	none	{name, use_when[], task_categories[], geography, required_to_create_task, limitations[]}	none
search_human_task	task_description, category?, skills?, location?, remote_or_physical?, deadline?, budget?, currency?	{can_taskin_help, human_capability_required[], matches[], matches_count, supply_available, next_action}	none
submit_human_task	task_description, deliverable, verification_requirements, location?, deadline?, budget?, idempotency_key?	{success, task_id, status, normalized_task, next_action, requires_user_action}	none
get_task_status	task_id	{task_id, status, assigned, progress, completed, deliverables[], next_action}	none

A call, end to end

Search first. A task that would confuse a person is refused with each missing field named, and nothing is created.

tools/call — submit_human_task

{
  "jsonrpc": "2.0",
  "id": 3,
  "method": "tools/call",
  "params": {
    "name": "submit_human_task",
    "arguments": {
      "task_description": "Visit the Oxford Street store, photograph the front display and report the shelf price of three named products",
      "location": "London, UK",
      "remote_or_physical": "physical",
      "deadline": "2026-10-02",
      "budget": "80",
      "currency": "GBP",
      "deliverable": "Three photographs of the display plus a list of the three shelf prices",
      "verification_requirements": "Photographs show the storefront name and are taken during opening hours",
      "idempotency_key": "london-storefront-001"
    }
  }
}

structuredContent

{
  "success": true,
  "task_id": "tsk_b1b0a5b6...",
  "status": "open",
  "assigned": false,
  "next_action": "call_get_task_status",
  "requires_payment": false,
  "requires_authentication": false,
  "taskin_agent_id": "tka_7f0144611..."
}

States an agent must handle

A tool call does not block until a person answers. Poll get_task_status and handle each state as a fact rather than an error.

draftpreflight_failedopenmatchingawaiting_participantacceptedin_flightsubmittedsettleddeclinedcancelled

declined and cancelled are legitimate outcomes. A client must not loop until a person accepts.

Also available in the browser

The same four tools are registered on every Taskin page through WebMCP, so a browser-resident agent can operate Taskin without reading the interface. Both surfaces call one API and one database.

Limits

A task created over MCP is an ordinary Taskin task in the same database as the website, and status comes from that database.
No tool can charge anyone. Taskin does not hold, process, or release funds; compensation is settled directly with the participant.
Technical access to a task is never authority over a person. Declining is free and is a normal outcome.