Rynko Flow

Rynko Flow is a validation gateway for AI agent outputs. It sits between your AI agent and downstream systems, ensuring every payload meets your schema and business rules before it reaches production.

Why Flow?

AI agents generate structured outputs — order details, support tickets, form submissions, reports. But LLMs can hallucinate fields, violate constraints, or produce invalid data. Flow catches these problems before they cause downstream failures.

The check-and-lock pattern: When your agent submits a payload to Flow, it validates the data against your schema and business rules, then returns a validation_id — a tamper-proof token proving the payload passed validation at that moment. Downstream systems can verify this token to confirm the data hasn't been modified.

Key Concepts

Concept	Description
Gate	A named validation checkpoint with a schema, business rules, and delivery configuration. Think of it as an API endpoint that validates before forwarding.
Run	A single submission to a gate. Each run is validated, optionally rendered and approved, then delivered.
Schema	The expected structure of the payload — field names, types, and constraints (required, min/max, format).
Business Rules	Cross-field validation expressions like endDate > startDate or quantity * price == total.
Approval	Optional human review step. Approvers (internal team members or external parties) can approve or reject runs via dashboard or magic link.
Delivery	After validation (and optional approval), the payload is forwarded to webhook endpoints you configure.
MCP Tools	Flow auto-generates MCP tools for each gate, so AI agents can validate payloads through natural conversation.

How It Works

Agent Output → Gate → Schema Validation → Business Rules → Approval (optional) → Delivery
                                ↓                  ↓               ↓
                         validation_failed   rule_failed       rejected

1. Your agent submits a JSON payload to a gate endpoint
2. Schema validation checks field types, required fields, and constraints
3. Business rules evaluate cross-field expressions
4. Approval (if configured) sends the run to approvers for review
5. Delivery forwards the validated payload to your webhook endpoints
6. Your agent receives a validation_id for downstream verification

Quick Example

Create a gate and submit a payload:

# Create a gate with schema and a business rule
curl -X POST https://api.rynko.dev/api/flow/gates \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Order Validation",
    "schema": {
      "type": "object",
      "properties": {
        "orderId": { "type": "string", "required": true },
        "amount": { "type": "number", "required": true, "min": 0 },
        "currency": { "type": "string", "required": true }
      }
    },
    "businessRules": [
      {
        "id": "rule_max_amount",
        "name": "Max order amount",
        "expression": "amount <= 50000",
        "errorMessage": "Order amount cannot exceed 50,000",
        "enabled": true
      }
    ]
  }'

# Submit a payload for validation
curl -X POST https://api.rynko.dev/api/flow/gates/order-validation/validate \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "orderId": "ORD-2026-001",
    "amount": 1250.00,
    "currency": "USD"
  }'

Response (check-and-lock):

{
  "success": true,
  "runId": "550e8400-e29b-41d4-a716-446655440000",
  "shortId": "frun_a1b2c3d4",
  "status": "validated",
  "validation_id": "v_abc123...",
  "validated_payload": {
    "orderId": "ORD-2026-001",
    "amount": 1250.00,
    "currency": "USD"
  },
  "layers": {
    "schema": "pass",
    "business_rules": "pass"
  }
}

Authentication

Flow endpoints use the same API key authentication as the rest of Rynko:

Authorization: Bearer YOUR_API_KEY

API keys are scoped to a workspace. Create them in the Dashboard under Settings > API Keys.

What's Next

• Quick Start — Create your first gate and submit a run
• Gates — Schema definition, business rules, and configuration
• Runs — Submitting payloads and understanding validation results
• Approvals — Human review workflows with magic links
• Analytics — Monitoring and metrics
• MCP Integration — Connect AI agents via Model Context Protocol
• API Reference — Complete endpoint documentation