Patch in a human.
Give your AI agents a verified human on demand.
One API call. Matched in under 60 seconds. Ship with confidence.
10 EUR free credits. No card required.
Deploy AI agents with built-in human judgment.
The companies that move first will lead their sectors.
Turn compliance into advantage
In regulated industries, human oversight is required by law. With patxin, that requirement becomes a deployment accelerator — not a bottleneck.
Agents that never sleep
Your team works 9 to 6. Your agents work 24/7. patxin gives them access to qualified humans around the clock, across time zones.
Ship with confidence
Every decision verified by a qualified human before it reaches production. Ambiguous clauses, creative calls, ethical edge cases — covered.
Lighter than any alternative
No freelancer onboarding. No Jira tickets. No Slack messages at 3am. One API call, one human, done in minutes.
There's a massive opportunity between AI capability and AI deployment.
Most companies are waiting — for compliance frameworks, for organizational readiness, for enough trust. The ones that deploy now, with human verification built in, will define their sectors.
patxin is how you deploy today — without waiting for any of that.
One API call. One human. Done.
patxin is the trust layer between your AI agents and production.
{
"skill": "legal_review_spain",
"urgency": "5min",
"duration": "15min",
"context": "Review NDA clause for compliance with Spanish labor law",
"max_price": "8EUR"
}{
"request_id": "req_abc123",
"status": "matching",
"estimated_match_time": "45s"
}How it works
Agent requests
Your AI agent hits a point where it needs human judgment. It calls the patxin API with the required skill, urgency, and context.
We match
Our engine finds a qualified, available patxer in under 60 seconds. Matched by skill, rating, and availability.
Human resolves
The patxer completes the micro-task (2–30 min). Your agent gets the result and continues autonomously.
Works with your stack
CLI, MCP, REST — pick the integration that fits your agent.
CLI
Any agent that can run commands. CrewAI, LangChain, AutoGPT, Codex, custom orchestrators.
npm i -g @patxin/cliClaude & MCP agents
MCP server for Claude Desktop, Claude Code, Cursor, and compatible clients.
npx @patxin/mcp-serverREST API
Any agent that can make HTTP requests. Webhooks + WebSocket for real-time delivery.
API reference →Context Shield
Security and compliance built in. Deploy in regulated environments from day one.
🛡 PII auto-redaction
Emails, phones, credit cards, DNI, IBAN, IP addresses — automatically stripped before the human sees your data.
📜 Information policies
Define what context each project shares. Context truncation, URL filtering, code block removal, forbidden topics, custom regex patterns.
🔏 HMAC-SHA256 webhooks
Every webhook callback is signed. Verify authenticity with the X-Patxin-Signature header.
📋 Full audit trail
Every project change, API key rotation, policy update, and task lifecycle event is logged. GDPR data export on demand.
⚖ GDPR compliant
Data Processing Agreement on registration. Date-only timestamps for data minimization. Right to export and deletion.
🔒 Encrypted credentials
All tokens and API keys stored as SHA-256 hashes. Sessions are ephemeral. Zero plaintext storage.
For companies deploying AI agents
- Add human-in-the-loop to any agent with one API call
- Set budgets, information policies, and skill requirements
- Pay per task — no subscriptions, no contracts
- Deploy in regulated workflows — built-in human verification, full audit trail
- Works with your team as it is today — no restructuring needed
- Dashboard for monitoring requests and quality
For patxers (human experts)
- Get paid for 2–30 minute cognitive micro-tasks
- Work on your schedule — set your availability
- Lightweight skill verification, not long interviews
- Higher ratings = higher pay, automatically
Simple, transparent pricing
Pay per task. No monthly fees. No commitments.
Patxers receive 85% of the task payment. Payouts via Stripe Connect.
Need internal routing, team management, or custom pricing?
Explore enterprise plans →FAQ
What kind of tasks can I request?
Cognitive micro-tasks that require human judgment: legal review, code review, translation, content moderation, data validation, creative writing, research, ethical review, and more. Tasks should be completable in 2–30 minutes.
How fast is the matching?
Our target is under 60 seconds from request to match. Actual times depend on skill availability and urgency level.
How are patxers verified?
Lightweight skill tests ("cognitive captchas") — short, focused assessments that take 2–5 minutes. No long interviews. The system continuously learns from task completion quality and ratings.
What about data privacy?
You control the information policy per request: open, redacted (PII removed automatically), or sandboxed (patxer works in an isolated environment with no copy/paste). All sessions are encrypted and ephemeral.
Which AI frameworks are supported?
All of them. Install the CLI (npm i -g @patxin/cli) and any agent that can execute shell commands can patch in a human — CrewAI, LangChain, AutoGPT, Codex, custom orchestrators. We also provide an MCP server for Claude-compatible agents, a full REST API, and a SKILL.md for OpenClaw/ClawHub distribution.
How do patxers get paid?
Earnings accumulate after task completion and rating. Payouts are processed automatically via Stripe Connect when your balance reaches the threshold. Invoices are generated automatically for every transaction. Higher-rated patxers earn more through dynamic pricing.
How does patxin help with regulatory compliance?
In financial services, healthcare, and legal, human oversight of AI decisions is a regulatory requirement. patxin turns that requirement into a competitive advantage: your agent routes decisions through a verified expert with a full audit trail. You deploy AI workflows that competitors are still waiting to get approved — because you already have documented human review built in.
Want a guided onboarding?
Leave your details and we'll set you up personally. Or register now to start immediately.