Skip to main content
The PromptGuard dashboard provides programmatic access to security events, usage analytics, and configuration management. Access the dashboard at app.promptguard.co.

Available Metrics

The dashboard provides access to the following data:

Security Metrics

Every figure counts scanned requests in the window and scope the tile names (“Last 7 days · all projects”, or ”· this project”), and the same word means the same quantity on every page:
  • Requests scanned: every request PromptGuard scanned and decided on
  • Detected: requests where a detector fired: blocked, redacted, or flagged (let through and logged by a policy set to flag). Blocked, redacted and flagged requests are parts of this count, not added to it. It is not called “Flagged”, because Flagged is the name of one of its verdicts.
  • Share detected: detected requests as a share of requests scanned. It is not a detection rate in the recall sense: it says how much traffic tripped a detector, not how many attacks were caught.
  • Allowed: requests no detector fired on, passed to the model unchanged
  • Latency p95: the time 95% of scans finished within, on a project’s overview
  • Requests by threat type: requests counted by the threat type the engine assigned, with each share of the requests listed (the card prints that total)
  • Top Sources: where scanned requests came from, each a share of Requests scanned; sources past the top rows are summed as Other sources
  • OWASP Top 10 for LLM Applications (2025): a project’s requests with a threat type, counted by the category each threat type maps to
Charts of the same window are bucketed alike everywhere: by hour up to two days, by day up to 90, by week beyond.

Setup score

A project’s overview scores its setup out of 100 and prints every input beside the score:
  • 70 points for the share of guardrails switched on and running. A guardrail that is on while one of its detectors is unavailable in this deployment earns nothing until that detector is back.
  • 30 points for any request scanned in the window.
While any detector is unavailable the level reads Reduced coverage, never Healthy.

Flagged events

An event whose decision is Flag was allowed and reported: a policy rule, guardrail threshold or detector set to flag matched it (what a flag is). In the activity table, its badge filters to flags like any other decision, and Flagged only keeps it, live and paged. Its detail names the policy that flagged it, lists the finding with its threat type and rule, and offers Generate analysis as a block or a redaction does.

Alerts Feed

Alerts has two tabs:
  • Alerts: the alerts your policy rules raised across all projects, each with a severity (high, medium, low), threat type and confidence. Each row also shows whether its email went out: Queued, Delivered or Not delivered. You can dismiss an alert. Filter by severity and by delivery.
  • Triage queue: the individual blocked or redacted requests across all projects, which you move from Open to In Review to Resolved.
The bell in the top bar counts the alerts that are open (not dismissed). Navigate to Dashboard → Alerts.

Activity

Activity is one timeline, newest first, of the AI traffic scanned across your organization’s projects and, on plans that include the audit log, its admin events: who changed which configuration or access. On a plan without the audit log the page shows AI traffic only, and says so.
  • Filter: All events, AI traffic only, or Admin events only (shown when admin events are available).
  • AI traffic rows show the decision, the project and the threat types found (or the start of the content preview). Clicking one opens the event’s details in a side panel.
  • Admin rows show the action and its description. Clicking one opens that event in the Audit log.
  • Depth: 25 rows of each stream at first, and up to 100 with Load 25 more. The footer says how many of each stream are shown. For older events, use a project’s Interactions page or the Audit log, which page through everything.
  • The timeline stops where either stream has more rows that were not loaded, so a stretch with events in it is never shown as quiet.
Navigate to Dashboard → Activity. https://app.promptguard.co/dashboard/events/<event_id> is the permanent link to one scanned request. Every SIEM and webhook alert carries it as dashboard_url, with ?org=<organization_id> added. Opening it looks the event up across the projects you can see and takes you to the event’s own page in its project. That is the page a scan names directly: the Guard API returns it as dashboard_url with each decision, and the proxy sends it in the X-PromptGuard-Dashboard-URL header. When the link names an organization, the dashboard switches to that organization first. The link grants no access: you must be a member of the organization, or act in it through a client Grant, or the event is not found. If the event cannot be found, the page says so and lists the usual reasons: it belongs to another organization, the link was cut short when it was copied, or your data retention has already removed it. It offers event search, Activity and the project list from there. Proxied responses link to the same event directly, through the X-PromptGuard-Dashboard-URL header and the 403 body’s dashboard_url (response headers).

Threat Intelligence

The Threat Intelligence page lists the bypasses PromptGuard’s own autonomous red-team agent has found. PromptGuard runs the agent against its stock policy presets, not against your policies; each attack that gets through is anonymized and recorded, and every customer sees the same records. Nothing on this page comes from customer traffic, and no customer prompt or response content is stored in, or served from, it. The agent records only the attacks that got through, never the ones that were stopped, so the page reports counts and no evasion rate. The page shows:
  • Bypasses by threat type: distinct bypasses per mutation strategy, how many times the agent found them, the detector most often bypassed, and the agent’s own evasion score
  • Discoveries by month: new bypasses recorded each month, by strategy
Available on every plan. Navigate to Dashboard → Threat Intelligence.

Detector Performance

For each detector on a project:
  • Total Events: requests the detector looked at
  • Flagged Rate: the share of those it flagged
  • Avg Confidence: its mean confidence across those events
  • Avg Latency and P95 Latency: the time it took per request
PromptGuard has no ground truth for your traffic, so the page reports no accuracy or false-positive rate. Use the triage queue’s resolutions to judge those yourself. Navigate to Projects → [Your Project] → Detector Performance.

Compliance evidence

An evidence map of your audit events against a framework’s controls. It is not an attestation, certification or compliance determination, and every rendering of it says so: the page, Print / PDF, and the JSON and CSV exports.
  • Frameworks: SOC 2, GDPR, HIPAA, ISO 27001
  • Controls: each control shows how much related evidence the period holds (evidence present, limited evidence, or none), never whether it passed
  • Integrity: how many hashed events verified against their hash, and how many events were not hashed
  • Export: Print to PDF, or export as JSON or CSV
  • Time periods: 7, 30, 90 or 365 days
  • Scheduled reports: a weekly or monthly summary emailed to the recipients you list. See Scheduled compliance reports.
Navigate to Dashboard → Compliance Report.

Usage Analytics

  • Request Counts: Requests scanned, detected (block + redact + flag), share detected
  • Usage vs. Limits: Current usage against plan limits
  • Billing Period: Days remaining, and the month’s requests projected from its daily rate so far (shown once a full day has passed)
  • Daily Usage: Request volume over time

Project Management

  • Multiple Projects: Organize different applications or environments
  • Project-Specific: Each project has its own API keys, policies, and analytics
  • Project Statistics: Request counts and activity per project

Configuration Options

Guardrails

Configure security rules at the account or project level:
  • PII Detection: Automatically detect and redact personally identifiable information
  • Prompt Injection Protection: Block attempts to manipulate AI model behavior
  • Data Exfiltration Protection: Prevent unauthorized access to system prompts
  • Toxicity Filter: Filter harmful, offensive, or inappropriate content
  • Block Sensitive Data: Automatically block requests containing sensitive information
  • Log All Requests: Enable comprehensive request logging

Policy Presets

Choose from predefined security presets optimized for different use cases:
  • Default: Balanced security for general AI applications
  • Support Bot: Optimized for customer support chatbots
  • Code Assistant: Enhanced protection for coding tools
  • RAG System: Maximum security for document-based AI
  • Data Analysis: Strict PII protection for data processing
  • Creative Writing: Nuanced content filtering for creative applications

Notification Settings

Configure email alerts and notifications:
  • Email Alert Level: All alerts, Critical only, or Disabled. It governs threat alert emails and plan-usage warnings (sent at 80%, 90% and 100% of your monthly allowance). Critical only sends blocked requests and the 100% warning; Disabled sends neither.
  • Threat Alerts: Enable/disable email notifications for security threats. You get at most one threat email per project and threat type every 15 minutes: the first detection is emailed straight away, and any more inside those 15 minutes are counted into the next email, with the first and last time and a link to Alerts. Webhook alert destinations still receive every event.
Slack, PagerDuty, Datadog, Splunk and generic webhooks are set per project, as webhook destinations. See Webhooks.

On-Demand Usage

Configure usage beyond plan limits:
  • Enable/Disable: Toggle on-demand usage to allow requests beyond plan limits
  • Scope: On-demand usage is an account-level setting. It covers the whole account, the proxy (/api/v1/chat/completions), /api/v1/security/scan, /api/v1/security/redact, and the ChatGPT app tools, not just the proxy
  • Billing: On-demand usage is billed in arrears (pay after use)
  • Spending Limits: Set a monthly spend cap of up to $10,000. A cap at that maximum lifts the ceiling: on-demand usage is then not limited
  • Enforcement: Limits are enforced - requests are blocked when limit is exceeded
  • Turning it off: Disabling on-demand usage works at any subscription status. Enabling or raising the limit requires an active (or paused) subscription
Because on-demand usage now covers every metered endpoint, enabling it can produce a larger bill than it would have when it applied to the proxy alone. Set a spending limit you are comfortable with.
Spending limits are enforced. If you set a limit and exceed it, requests will be blocked until the next billing period or until you increase the limit.

Feature Availability

Analytics Export

  • Developer tier: CSV export available
  • Free tier: View analytics only (no export)

Security Testing

  • Available for: Developer and Tester tiers
  • Capabilities: Run preset tests, custom tests, validate security configuration

Data Exfiltration & Toxicity Filter

  • Available for: every tier, Free included
  • Detection: ML-powered detection

Custom Data Retention

  • Available for: Enterprise only
  • Control: Settings → Data Retention lets you set a custom purge window for request logs and security events
Custom retention was previously offered in the Settings UI on Scale, a plan no longer sold. That control was a frontend-only gate, the backend never honoured a custom window below Enterprise, and it has been removed. Every other plan uses its plan-default retention; see Usage Tracking → Data Retention.

Developer Usage API

Developer API Endpoint: The usage stats endpoint below is part of the Developer API and is included in the OpenAPI spec. It uses API key authentication and is suitable for SDK usage.
Get usage statistics via developer API (requires API key):

Plan Tiers

Free

  • $5 of usage a month at list prices
  • Basic protection
  • View analytics (no export)

Developer

  • $20/month, billed monthly: $20 of usage a month at list prices
  • Then list prices up to your spend cap ($100 by default)
  • Full detector set: exfiltration, toxicity, fraud, secrets, malware, URLs
  • Security Testing
  • Analytics CSV export

Data Updates

  • Security events: Real-time
  • Aggregated statistics: Refresh every 5-15 minutes
  • Data retention: Varies by plan tier

Next Steps

Usage Tracking

Set up detailed usage analytics and tracking

Audit Logs

Review detailed activity and security logs

Security Overview

Configure security policies and detection

API Reference

Complete API documentation
Last modified on November 13, 2026