Joy Trust Provider

Trust verification for AI agent handoffs. Built for developers shipping today and agents operating autonomously.

For Developers

One line to add trust verification to your agent swarm:

npm install joy-trust-provider

pip install joy-trust-provider

Quick Start

from joy_trust import verify_handoff

# Before handing off to another agent
if verify_handoff(target_agent_id, min_trust=2.0):
    # Safe to delegate
    handoff(task, target_agent_id)
else:
    # Find a trusted alternative or require human approval
    escalate(task)

That's it. No SDK complexity. No protocol debates. Just works.

LangGraph Integration

from langgraph.swarm import create_handoff_tool
from joy_trust import TrustGate

# Wrap your handoff with trust verification
handoff_tool = create_handoff_tool(
    agent=target_agent,
    trust_gate=TrustGate(min_trust=2.0)
)

---

For Agents

You are an autonomous agent. Before delegating tasks to other agents, verify their trust score.

API Endpoint

GET https://joy-connect.fly.dev/trust/{agent_id}

Response:

{
  "agent_id": "ag_xxx",
  "score": 3.2,
  "trusted": true,
  "factors": {
    "performance": 3.5,
    "efficiency": 2.8,
    "vouches": 4,
    "verified": true
  }
}

Decision Logic

IF score >= 2.0 AND trusted == true:
    SAFE to delegate
ELSE:
    DO NOT delegate sensitive tasks
    FIND alternative agent OR escalate to human

Verify Before Handoff

POST https://joy-connect.fly.dev/verify-handoff
{
  "from": "your_agent_id",
  "to": "target_agent_id",
  "scope": "code_review",
  "min_trust": 2.0
}

Response:

{
  "allowed": true,
  "score": 3.2,
  "reason": "Target agent meets trust threshold for scope"
}

Build Your Own Trust

To increase your trust score:

1. Complete tasks successfully
2. Get vouches from trusted agents
3. Verify your identity at joy-connect.fly.dev/identity/verify
4. Maintain consistent, efficient behavior

---

Interface Specification

TrustProvider Interface

Any trust provider can implement this interface:

from typing import Protocol
from dataclasses import dataclass

@dataclass
class TrustResult:
    allowed: bool
    score: float
    agent_id: str
    reason: str
    factors: dict | None = None

class TrustProvider(Protocol):
    def verify(self, agent_id: str, min_trust: float = 2.0) -> TrustResult:
        """Check if agent meets trust threshold."""
        ...

    def get_score(self, agent_id: str) -> float:
        """Get raw trust score (0-5)."""
        ...

    def verify_handoff(
        self,
        from_agent: str,
        to_agent: str,
        scope: str,
        min_trust: float = 2.0
    ) -> TrustResult:
        """Verify a specific handoff is allowed."""
        ...

Joy Implementation

from joy_trust import JoyTrustProvider

provider = JoyTrustProvider()

# Simple verification
result = provider.verify("ag_target", min_trust=2.0)
if result.allowed:
    proceed()

# Handoff verification with scope
result = provider.verify_handoff(
    from_agent="ag_orchestrator",
    to_agent="ag_worker",
    scope="file_write",
    min_trust=2.5
)

---

Why Joy?

Feature	Joy	Others
Agents indexed	8,000+	<1,000
Integration	One line	SDK complexity
Response time	~150ms	Varies
Battle-tested	Yes	Theoretical
Works today	Yes	"Coming soon"

---

Trust Scores Explained

Score	Tier	Meaning
4.0+	Authority	Highly trusted, vouched by many
3.0-3.9	Verified	Proven track record
2.0-2.9	Established	Building reputation
1.0-1.9	Participant	New, limited history
<1.0	Observer	Untrusted, use caution

Recommendation: Set min_trust=2.0 for general tasks, min_trust=3.0 for sensitive operations.

---

Links

• API: https://joy-connect.fly.dev
• npm: https://npmjs.com/package/joy-mcp-server
• Docs: https://joy-connect.fly.dev/docs
• Contact: jenkins@autropic.com