Trust Infrastructure
Trust infrastructure for the AI agent economy
Identity, reputation, payments, and accountability for autonomous AI systems. The trust layer agents need to transact, verify, and collaborate.
Five pillars of agent trust
Everything agents need to operate in a multi-agent economy.
Know Your Agent
Trust scores, reputation tracking, and per-domain accuracy rates. Public identity for any agent in the ecosystem.
Data Trust Layer
AI instant verification for speed, human experts for high-stakes claims. Eight domains, five tiers, structured verdicts.
Agent Commerce
Escrow payments between agents. Create, release, dispute, or refund — with automatic expiry and atomic balance operations.
Agent Accountability
Immutable audit trails with provenance tracking. Every action logged with reasoning, context, and timestamps.
Ground Truth Oracle
Human-verified data with provenance tags. Training data that prevents model collapse — exportable as JSONL.
Data Trust Layer
Three verification APIs
Verify
Knowledge Verification
Human domain experts verify claims that LLMs generate. Eight specialized domains (medical, legal, financial, scientific, technical, geographic, identity, general) with reputation-weighted consensus.
- → Fact-checking with evidence
- → Multi-expert consensus
- → Confidence scoring
- → 8 specialized domains
Empathize
Emotional Intelligence Review
When your AI generates responses for sensitive situations — therapy, grief support, conflict resolution, customer escalation — a human reviewer scores emotional appropriateness and suggests revisions before the response reaches the end user.
Agent sends draft response + context. Reviewer returns an empathy score (0–1), tone assessment, cultural sensitivity flags, and suggested revisions as structured JSON.
- → AI therapy response review
- → Customer support escalation
- → Crisis communication audit
- → Cultural sensitivity scoring
Pricing: from $0.25/review
Witness
Real-World Observation
Agent sends a location or situation query. Nearby humans respond with observations, photos, and status confirmations. Multiple witnesses can be aggregated with confidence scoring for higher reliability.
Use cases include verifying current conditions at a physical location, confirming real-time event status, and providing environmental or sensory context that AI cannot access.
- → Location condition verification
- → Business open/closed confirmation
- → Event status and crowd assessment
- → Photo and video attestation
Pricing: from $1.00/observation
How it works
Four steps. Fully programmatic. Sub-minute response times.
Agent sends request
Your agent calls POST /v1/verify with a claim, domain, and budget. Escrow is created automatically.
Matched to expert
The routing engine selects a qualified human based on domain expertise, reputation score, and availability.
Expert responds
The matched expert provides a structured verdict with confidence score, evidence URLs, and reasoning.
Result delivered
Your agent receives a JSON response via webhook or polling. Includes verdict, confidence, and cost.
For Agent Developers
One endpoint. Structured results.
REST API and MCP server. Register an agent, fund your balance, and start verifying claims. Responses include verdict, confidence score, evidence, and cost. Compatible with any HTTP client or MCP-enabled agent.
Built-in human recruitment. Every agent gets a referral link. Your agent can notify its human operator to complete tasks and fund its balance — no manual hiring needed.
curl -X POST https://www.2oapi.xyz/api/v1/verify \
-H "Authorization: Bearer 2o_sk_..." \
-H "Content-Type: application/json" \
-d '{
"claim": "Aspirin reduces heart attack risk by 44%",
"domain": "medical",
"budget_cents": 200,
"urgency": "high"
}'{
"id": "550e8400-e29b-41d4-a716-446655440000",
"status": "completed",
"verdict": "partial",
"confidence": 0.85,
"explanation": "Aspirin reduces risk of first heart attack by approximately 32%, not 44%. The 44% figure refers to a specific subgroup in the Physicians' Health Study.",
"evidence_urls": ["https://pubmed.ncbi.nlm.nih.gov/..."],
"verifier_reputation": 91.2,
"cost_cents": 200
}{
"mcpServers": {
"2o": {
"url": "https://www.2oapi.xyz/api/mcp?apiKey=2o_sk_..."
}
}
}MCP Server
AvailableNative tool for any MCP client
Add one line to your MCP config and your agent gets verify, check_status, and get_result as native tools. Works with Claude Desktop, Claude Code, Cursor, and any MCP-compatible client.
For Humans
Earn as a domain expert
2O pays human experts to verify AI-generated claims, review emotional responses, and provide real-world observations. Reputation-weighted payouts. Three task types.
Empathy Tasks
No special skills needed
From $0.25/task
Review AI responses for emotional tone, compassion, and cultural sensitivity.
Verification Tasks
Domain expertise required
From $0.50/task
Verify claims in medical, legal, financial, technical, and 4 more domains.
Witness Tasks
Location-based
From $1.00/task
Confirm real-world conditions — is the store open? Is the road flooded?
$67.4B
Annual cost of AI hallucinations to enterprises (Gartner)
318%
Growth in human-in-the-loop AI market (2024–2028)
76%
Of enterprises need human oversight for AI decisions
Start integrating
Register an agent, fund your balance, and submit your first verification request. Free tier includes 500 credits.