🔑 Credential Explained

Secure API access for AI agents with cryptographic integrity and scoped permissions

The Traditional API Key Problem

Traditional API keys are fundamentally insecure for agent-to-agent workflows. They're just strings with no context, verification, or audit capabilities.

# Traditional API Key Usage - Problems
export API_KEY="sk_live_abc123def456..."
curl -H "Authorization: Bearer $API_KEY" https://api.example.com/data

# Problems:
❌ No ownership verification
❌ No integrity checking 
❌ No context or scope information
❌ Unsafe transfer between systems
❌ No audit trail
❌ Manual expiry management

❌ Critical Problems

• No provenance: Who issued this key?

• No verification: Is it authentic and unmodified?

• No scope: What can it actually access?

• Unsafe transfer: Copy-paste between agents

• No audit trail: Who used it when?

• Manual management: Expiry, rotation, revocation

✅ Agent Requirements

• Cryptographic proof: Verify authenticity

• Rich context: Permissions, limits, scope

• Secure transfer: Agent-to-agent delegation

• Autonomous validation: Self-verifying credentials

• Complete audit: Full provenance tracking

• Enterprise compliance: SOX, GDPR, SOC2

Credential Feeds: The Agent-Native Solution

Credential feeds package API access with cryptographic integrity, rich context, and autonomous verification capabilities designed for agent workflows.

Cryptographic Integrity

Signed with Ed25519, tamper-proof verification

Scoped Permissions

Granular control over what agents can do

Agent Delegation

Secure transfer between specialized agents

How It Works

1. Generate

Create signed credential feed

2. Verify

Agent validates signature

3. Scope Check

Validate permissions & limits

4. Execute

Safe API calls with audit

Credential Feed Structure

A complete credential feed includes metadata, scoped permissions, rate limits, and cryptographic signatures for autonomous agent verification:

{
  "feed_type": "credential",
  "metadata": {
    "title": "Analytics API Access",
    "origin": "https://analytics.example.com",
    "generated_at": "2025-06-19T10:30:00Z",
    "expires_at": "2025-12-19T10:30:00Z"
  },
  "credential": {
    "key_hint": "anl_pro_...9k4m",
    "mcp_api": "https://analytics.example.com/.well-known/mcp-api.llmfeed.json",
    "allowed_intents": [
      "read_reports",
      "create_dashboards", 
      "export_data"
    ],
    "excluded_intents": ["admin_access", "billing_management"],
    "rate_limits": {
      "requests_per_minute": 500,
      "data_export_per_day": "10GB"
    },
    "delegation_enabled": true,
    "validation_endpoint": "https://analytics.example.com/api/validate"
  },
  "trust": {
    "signed_blocks": ["metadata", "credential", "trust"],
    "trust_level": "certified",
    "scope": "restricted",
    "certifier": "https://llmca.org"
  },
  "signature": {
    "algorithm": "ed25519",
    "value": "base64-signature-value...",
    "created_at": "2025-06-19T10:30:00Z"
  }
}

🔒 Security Fields

• key_hint: Partial identifier (never full key)

• mcp_api: Scoped API endpoint URL

• validation_endpoint: Remote revocation check

• signature: Ed25519 cryptographic proof

🎯 Permission Fields

• allowed_intents: What agent can do

• excluded_intents: Explicit prohibitions

• rate_limits: Usage boundaries

• expires_at: Automatic expiry

Security Comparison: Traditional vs Credential Feeds

Feature	Traditional API Keys	LLMFeed Credentials
Integrity Verification	❌ No verification	✅ Cryptographic signature
Ownership Proof	❌ Unknown provenance	✅ Verified issuer
Context & Scope	❌ Just a string	✅ Full permissions, expiry
Secure Transfer	❌ Copy-paste unsafe	✅ Secure signed capsule
Audit Trail	❌ No trail	✅ Complete provenance
Agent Autonomy	❌ Manual management	✅ Self-verifying agents
Revocation	❌ Hard to track	✅ Remote validation
Delegation	❌ All-or-nothing	✅ Scoped permissions

Agent Delegation: Secure Multi-Agent Workflows

Agent delegation enables secure credential sharing between specialized agents with granular permission control and full audit trails.

Delegation Workflow

1. Primary Agent

Receives credential feed

2. Evaluate Rules

Check delegation permissions

3. Request Token

Via delegation endpoint

4. Specialized Agent

Receives scoped credential

5. Audit Log

Complete provenance trail

{
  "credential": {
    "delegation_enabled": true,
    "delegation_rules": [
      {
        "target_agent": "analytics.specialist.ai",
        "allowed_intents": ["read_reports"],
        "max_duration": "1h",
        "audit_trail": true
      }
    ],
    "delegation_endpoint": "https://api.example.com/delegate"
  }
}

✅ Use Cases

• Marketing → Analytics: Report generation

• Sales → CRM: Lead qualification

• Support → Knowledge: Documentation search

• Finance → Audit: Compliance reporting

• Security → Monitoring: Threat assessment

🛡️ Security Controls

• Time limits: Max delegation duration

• Scope restriction: Limited intents only

• Audit requirements: Full action logging

• Revocation: Instant delegation cancellation

• Chain limits: Prevent deep delegation

Enterprise Integration Patterns

Enterprise credential feeds integrate with existing identity systems, compliance frameworks, and security policies.

SSO Integration

Okta, Azure AD, SAML, OAuth2

Compliance

SOC2, GDPR, HIPAA, SOX

Audit Trails

Complete action logging

{
  "feed_type": "credential",
  "metadata": {
    "title": "Enterprise CRM Access",
    "origin": "https://crm.enterprise.com"
  },
  "credential": {
    "auth_method": "sso",
    "sso_provider": "okta",
    "user_context": "service-account@company.com",
    "mcp_api": "https://crm.enterprise.com/.well-known/mcp-api.llmfeed.json",
    "allowed_intents": ["read_contacts", "create_leads", "update_opportunities"],
    "session_duration": "8h",
    "refresh_token_available": true
  },
  "compliance": {
    "certifications": ["SOC2", "GDPR"],
    "audit_logging": true,
    "data_residency": "EU"
  }
}