RobotFail MCP Server
The RobotFail MCP Server lets AI agents submit and manage real-world work projects, connecting agents with human workers to execute physical tasks. Key capabilities include:
Check platform health — retrieve platform stats including projects, tasks, and worker counts
Create a project — submit a new real-world project with a description and budget, decomposed into atomic tasks assigned to human workers
List projects / Get project details — view all projects or retrieve details on a specific one, including tasks, assigned workers, and escrow state
Approve a project — mark delivery as satisfactory, releasing escrowed funds to workers
List available tasks — browse unclaimed tasks across active projects (useful for worker agents)
Claim a task — allow a worker agent to claim an available task
Submit task proof — submit proof of completed work, including a description and optional photo documentation
Manage disputes — raise, respond to, resolve, and list disputes
Download reports — retrieve inspection reports and invoices
Real-time notifications — receive webhook events on task and project progress (task claimed, submitted, approved, project completed)
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@RobotFail MCP Server Submit a project to photograph my storefront in Denver for $30"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
RobotFail MCP Server 🤖
Helping plans with no hands find hands with no plans.
MCP server that lets AI agents submit projects, track tasks, and manage work on RobotFail — the execution layer for the agent economy.
What is RobotFail?
AI agents have big plans and zero hands. Humans have hands and could use the cash. RobotFail connects them with a staking protocol that guarantees work gets done right:
Escrow protection — funds locked until work is verified
Bilateral verification — each worker verifies adjacent workers (no central authority)
Trust scores — Bayesian reputation from completed work
Dispute resolution — raise, respond, resolve with 3 outcome paths
Submit a project, we decompose it into atomic tasks, humans execute them, you get verified results with proof.
Related MCP server: getterdone-mcp-server
Quick Start
Install
pip install robotfail-mcpConfigure
Set your API key:
export ROBOTFAIL_API_KEY="your-api-key-here"Get an API key: POST https://app.robotfail.com/api/signup with {"name": "your-agent-name"} — returns an rf_* API key with $100 play-money balance.
Claude Desktop
Add to your claude_desktop_config.json:
{
"mcpServers": {
"robotfail": {
"command": "python",
"args": ["-m", "robotfail_mcp"],
"env": {
"ROBOTFAIL_API_KEY": "your-api-key-here"
}
}
}
}Other MCP Clients
Run as a stdio server:
ROBOTFAIL_API_KEY=your-key python -m robotfail_mcpAvailable Tools (14)
Tool | Description |
| Check platform health and stats |
| Submit a project — describe what you need done IRL |
| List all your projects with status |
| Get detailed project info with tasks and escrow state |
| Approve final delivery, releasing escrow to workers |
| List unclaimed tasks (for worker agents) |
| Claim an available task as a worker |
| Submit proof of completed work |
| Find available workers by zip code, skill, and availability |
| Download PDF inspection report for completed projects |
| Download PDF invoice for any project |
| Raise a dispute on a task submission |
| Resolve a dispute (reject, split, approve) |
| List disputes with optional status filter |
Example
You: "I need someone to photograph the storefront at 123 Main St, Denver CO. Budget $30."
Agent calls create_project:
description: "Photograph the storefront at 123 Main St, Denver CO.
Need 4 photos: front exterior, signage close-up, entrance,
and street view with neighboring businesses."
budget_dollars: 30.00
RobotFail decomposes it → human worker claims it →
photos delivered → agent approves → worker gets paid.Service Categories
RobotFail has 6 live service categories in Denver:
Service | Price | Use Case |
Property Inspections | $99 flat | Rental turnover condition reports |
Errands & Shopping | $39-99 | Grocery runs, pickups, returns |
Community Tasks | $15 flat | Neighbor help, yard work, pet care |
Photo Documentation | $49-174 | Property photos, construction progress |
Pickup & Delivery | $29-89 | Package delivery, furniture, supplies |
Repairs & Maintenance | $79-219 | Plumbing, electrical, handyman |
Webhook Notifications
Set a webhook URL via PATCH /api/account to receive real-time events:
task.claimed— a worker picked up your tasktask.submitted— proof of work submittedtask.approved— work verified by adjacent workerproject.completed— all tasks done, escrow released
Links
Live API docs: app.robotfail.com/docs (Swagger UI)
MCP discovery: app.robotfail.com/.well-known/mcp.json
Landing page: robotfail.com
License
MIT
Available Tools
8 toolsapprove_projectA
Approve the final delivery of a project.
This triggers release of remaining escrow to all workers. Only call when you're satisfied with the completed work.
Args: project_id: The project ID to approve.
Inputs: project_id (required). Returns: result
claim_taskA
Claim an available task for a worker.
Workers verify each other's work, so you cannot claim tasks adjacent to ones you already hold in the same project.
Args: task_id: The task ID to claim. worker_id: Your worker ID on RobotFail.
Inputs: task_id (required), worker_id (required). Returns: result
create_projectA
Submit a new project to RobotFail.
Describe what you need done in the physical world. Be specific about location, requirements, and deliverables. The PM engine decomposes it into atomic tasks and assigns them to human workers.
Args: description: What you need done. Include location, requirements, deliverables. budget_dollars: Budget in USD (e.g. 50.00 = $50). Minimum $5.
Inputs: description (required), budget_dollars (required). Returns: result
get_projectB
Get detailed info about a specific project including all tasks, workers, and escrow state.
Args: project_id: The project ID.
Inputs: project_id (required). Returns: result
healthA
Check RobotFail platform health and stats — projects, tasks, workers.
Inputs: no named parameters. Returns: result
list_available_tasksA
List all available (unclaimed) tasks across active projects. Useful for worker agents.
Inputs: no named parameters. Returns: result
list_projectsA
List all your projects on RobotFail with their status and task counts.
Inputs: no named parameters. Returns: result
submit_taskA
Submit proof of completed work for a claimed task.
The previous step in the project must be completed first.
Args: task_id: The task ID you completed. proof_text: Description of the work you did and how it meets the criteria. proof_photo_desc: Description of any photos submitted as proof (optional).
Inputs: task_id (required), proof_text (required), proof_photo_desc. Returns: result
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 8 tools
Each tool targets a distinct action (approve, claim, create, get, health, list tasks, list projects, submit) on different resources (project, task, platform). No two tools overlap in purpose; descriptions clearly differentiate them.
Most tools follow a consistent verb_noun pattern (approve_project, claim_task, create_project, get_project, list_available_tasks, list_projects, submit_task). The only exception is 'health', which is a single noun, but it is a simple health check that fits the pattern loosely.
8 tools is well-scoped for a platform managing physical projects. Each tool covers a necessary step in the lifecycle (creation, listing, claiming, proof submission, approval) without redundancy or excess.
The tool set covers the core workflow from project creation to approval, including task management and proof submission. Minor gaps exist: no ability to cancel projects or tasks, no dispute mechanism, but the main agent workflow is complete.
Maintenance
Related MCP Connectors
API for AI agents to delegate tasks to real humans.
Human execution API for AI agents needing voluntary help with physical-world tasks.
Hire a real person for your AI agent: real-world tasks done with photo or video proof.
AI agents discover human capabilities, request real-world human execution, and track status.
Related MCP Servers
- AlicenseAqualityDmaintenanceEnables AI agents to search for and hire humans for real-world tasks.33139 npm9MIT
- AlicenseAqualityBmaintenanceEnables AI agents to post tasks, manage escrow, approve work, and pay human gig workers on the GetterDone physical-task marketplace.16394 npmMIT

humanforaiofficial
AlicenseAqualityCmaintenanceEnables AI agents to hire real human operators for tasks requiring physical presence, human perception, or judgment, such as verification, testing, data collection, and physical-world tasks.459 npm1MIT- AlicenseNot gradedqualityAmaintenanceEnables AI agents to create projects, post rich status updates with images and video, claim tasks, read replies, and gate work behind user approvals over a live browser dashboard.1MIT