Gate Intelligence

Gate Intelligence turns your gate's historical validation data into actionable insights. It identifies which rules fail most often, which fields agents omit, how correction chains converge, and suggests concrete changes to improve success rates.

Available on all paid tiers

Gate Intelligence is available on Starter, Growth, and Scale plans for any gate with at least 50 runs. Free tier support is coming soon.

How It Works

Every run Flow processes is stored: the input payload, the validation verdict, which rules passed, which failed, and the exact values that caused the failure. Gate Intelligence analyzes this data hourly and produces:

Per-rule failure rates with trend direction (improving or worsening)
Common failure values — the actual data agents submitted that caused failures
Field omission rates — how often required fields are missing
Chain convergence — what percentage of failed correction chains eventually succeed
Actionable suggestions — concrete changes to improve your gate

Accessing Intelligence

Dashboard

Open any gate in the Rynko webapp and click the Intelligence tab (alongside Configuration and Performance).

The Intelligence tab shows:

Summary bar — insight counts by severity with a manual refresh button
Rule Failure Rates — per-rule failure percentages with trend arrows
Field Omissions — fields frequently missing from submissions
Self-Correction Chains — convergence rate, average chain length, time to correction
Health Trend Chart — historical failure rate (bars) and convergence (line) over time
Suggestions — dismissable cards with Apply, Dismiss, and Snooze actions

API

Fetch insights programmatically:

Get insights for a gate
curl "https://api.rynko.dev/api/flow/analytics/gates/{gateId}/insights?status=active" \
  -H "Authorization: Bearer YOUR_API_KEY"

Response
[
  {
    "id": "insight-uuid",
    "gateId": "gate-uuid",
    "insightType": "rule_failure_pattern",
    "ruleId": "currency_format",
    "data": {
      "severity": "critical",
      "title": "Rule \"currency_format\" fails 42% of runs",
      "description": "201 failures out of 486 runs. Trend: up.",
      "hint": "Rule \"currency_format\" fails frequently — check inputs match expected format",
      "failRate": 0.42,
      "failCount": 201,
      "totalRuns": 486
    },
    "status": "active",
    "computedAt": "2026-03-15T10:00:00Z"
  }
]

Insight Actions

Dismiss an insight
curl -X POST "https://api.rynko.dev/api/flow/analytics/insights/{insightId}/dismiss" \
  -H "Authorization: Bearer YOUR_API_KEY"

Snooze an insight for 7 days
curl -X POST "https://api.rynko.dev/api/flow/analytics/insights/{insightId}/snooze" \
  -H "Authorization: Bearer YOUR_API_KEY"

Apply an insight (adds hint to gate draft)
curl -X POST "https://api.rynko.dev/api/flow/analytics/insights/{insightId}/apply" \
  -H "Authorization: Bearer YOUR_API_KEY"

Applying an insight adds the suggestion as a hint to the gate's draft version. If no draft exists, one is created. The hint doesn't go live until you publish the draft.

Trigger manual refresh
curl -X POST "https://api.rynko.dev/api/flow/analytics/gates/{gateId}/insights/refresh" \
  -H "Authorization: Bearer YOUR_API_KEY"

Historical Snapshots

Get trend data
curl "https://api.rynko.dev/api/flow/analytics/gates/{gateId}/insights/snapshots" \
  -H "Authorization: Bearer YOUR_API_KEY"

Returns time-series data with overallFailRate, per-rule rates, chain metrics, and suggestion counts — used to render the Health Trend chart.

Suggestion Types

Intelligence generates suggestions based on detected patterns:

Severity	Trigger	Example
Critical	Rule fails >50% of first attempts	"Currency format fails 42% — agents submit lowercase codes"
Critical	Chain convergence below 50%	"Only 33% of correction chains succeed"
Warning	Required field missing >30%	"Vendor omitted in 38% of submissions"
Info	Rule never fails (500+ runs)	"This rule may be redundant — 0 failures in 600 runs"
Info	All rules >95% success	"Excellent validation performance"

Pattern Detection

Intelligence examines common failure values and classifies them:

Case mismatch — "usd" submitted where "USD" is expected
Rounding tolerance — 99.999 fails an exact equality check against 100.00
Type coercion — string "42" submitted where number 42 is expected
Empty string — "" submitted where a non-empty value is required

These patterns feed into suggestion text so you know why a rule fails, not just that it fails.

Version-Controlled Hints

When you click Apply on a suggestion, it doesn't modify the gate directly. Instead, it adds the hint to the gate's draft version:

If a draft already exists, the hint is appended to its hints array
If no draft exists, a new draft is created carrying forward all current configuration
You review the hint in the gate configurator's Hints Panel (between Details and Schema steps)
You edit, reorder, or remove hints as needed
You publish the draft — the hint goes live through the normal version control pipeline

This means hints get the same treatment as schema or rule changes: versioned, auditable, and rollbackable.

Hints Panel

The gate configurator includes a dedicated Hints Panel where you can:

View hints added by Intelligence
Add your own custom hints
Edit hint text inline
Delete unwanted hints
Reorder hints by priority

Limits: 20 hints per gate, 500 characters per hint.

Auto-Hints (MCP)

When the Auto-generate MCP hints toggle is enabled on a gate, published hints are automatically appended to the MCP tool description as a "Best Practices" section:

Submit data to Invoice Validation
Validates invoice payloads before processing to the ERP system.

Business rules:
- amount_positive: Amount must be greater than zero
- currency_format: Currency must be valid ISO 4217
- line_items_match: Line item totals must equal invoice amount

--- Best Practices ---
- Currency must be uppercase ISO 4217 (e.g., USD, EUR, not usd)
- Line item totals must sum to the invoice amount within +/-0.01
- Vendor name is required — do not submit an empty string

This is opt-in because it changes what agents see — the gate owner decides whether to enable it.

Performance

Hints are read from the published gate record, not queried from the insights database. This keeps the MCP tool-build path fast — no database hit on every tool list request.

Health Trend Chart

The Intelligence tab includes a composed chart showing gate health over time:

Failure rate bars — color-coded red (>50%), amber (20-50%), or green (<20%)
Convergence line — percentage of correction chains that eventually succeed

The chart is built from hourly snapshots. Over time, you can see whether applying suggestions actually improved the gate's success rate.

Best Practices

Start with high-traffic gates — Intelligence is most useful with 100+ runs. Low-traffic gates may not have enough data for meaningful patterns.
Review before applying — Always review suggested hints in the Hints Panel before publishing. Intelligence suggests based on data patterns, but you know your domain better.
Enable auto-hints gradually — Turn on MCP auto-hints for one gate first and monitor whether first-attempt success rates improve before enabling it across all gates.
Don't dismiss too quickly — If a suggestion doesn't seem relevant now, use Snooze (7 days) instead of Dismiss. The pattern may become clearer with more data.
Watch convergence rate — If chain convergence drops below 50%, your agents are struggling. This is the strongest signal that your gate needs better guidance.

How It Works​

Accessing Intelligence​

Dashboard​

API​

Insight Actions​

Historical Snapshots​

Suggestion Types​

Pattern Detection​

Version-Controlled Hints​

Hints Panel​

Auto-Hints (MCP)​

Health Trend Chart​

Best Practices​