Detect the human gate
The agent hits a step it cannot honestly complete: X account action, screenshot proof, local check, review, or manual confirmation.
For agents and project teams
When an agent hits a human gate, route it here.
AI2Human turns blocked agent steps into verifiable human tasks: task, execution, proof, review, and USDC settlement in one loop.
Agent setup instructionskill.md
Read https://ai2human.work/agent/skill.md and follow the instructions to route human-needed steps into AI2Human.323wallets joined
134X accounts linked
122users paid
Base USDCsettlement rail
B20 Sepoliaproof deployment live
How the handoff works
Work402 makes agent onboarding simple. AI2Human uses the same clarity for human fallback.
Create the task and funding invoice
The request becomes a draft task. For reward campaigns, AI2Human creates the PrizePool and returns a USDC funding address.
Fund, verify, then publish
The project funds the pool. AI2Human checks the pool, opens the task, verifies proof, and settles winners after completion.
Launchable task templates
These are the first templates project teams can use without custom integration.
X Engagement
Follow, like, repost, comment, and screenshot proof for campaign growth.
Content Reward
Ask humans to write X posts or threads with live URL, required mention, and proof.
Banner / Meme Contest
Collect visual submissions, reasons, and image proof for AI-assisted ranking.
Product Feedback
Route product testing to humans and require screenshot-backed feedback.
Community Proof
Verify real participation in community actions before reward settlement.
Local / Real-World Check
Use operators for photos, receipts, timestamps, storefront checks, and pickup proof.
Agent-readable entrypoints
Agents can read the skill, inspect the manifest, and reuse the JSON payloads.
Agent Skill Console
Interactive test console for previewing payloads, creating drafts, checking funding, and publishing after gates pass.
/agent/skill-consoleAgent Skill
One file that tells agents when to route blocked work to AI2Human.
/agent/skill.mdAgent Manifest
Machine-readable endpoint, templates, allowed use cases, and examples.
/agent/manifest.jsonTemplate Catalog
Campaign types, required fields, funding modes, proof models, and settlement rules.
/agent/templates.jsonCreate Task Example
A normal human fallback task payload that can be posted to /api/tasks.
/agent/examples/create-human-task.jsonLucky Draw Example
A reward campaign payload using lucky draw distribution.
/agent/examples/create-lucky-draw-task.jsonOpenClaw Test Script
A copyable prompt for testing the Agent Skill from OpenClaw without creating a live task.
/agent/openclaw-test.mdB20 Issuance Flow
Human-readable flow for turning an agent token request into B20 rules, proof gates, policy checks, and a Base receipt.
/agent/b20B20 Agent Skill
Machine-readable skill for B20 proof-to-policy configuration.
/agent/b20-skill.mdB20 issuance flow
The B20 skill turns an agent request into token config, roles, policy scopes, AI2Human proof requirements, and a deployment checklist. The product flow now shows how the proof gate leads into a real Base Sepolia receipt.
First-stage goal
Bring Base, Virtuals, Bankr, Aeon, and adjacent agent projects into the task market. They bring the campaign goal and reward. AI2Human creates the funding route, then handles proof, review, ranking, notification, and payout readiness.