Skip to main content

Automation & AI

The Automation & AI page is the central configuration hub for WhiteOwl's AI-powered features and programmatic API access. It brings together AI model configuration, feature toggles, service connection health, MCP API key management, Claude Desktop integration, audit logging, and API documentation.

A link to the MCP Docs is available in the top-right corner for detailed API integration documentation.

AI Configuration

The top section controls all AI-powered capabilities in WhiteOwl.

AI Features Toggle

The master AI Features toggle enables or disables all AI-powered features across the platform. When disabled, WhiteOwl operates in a fully deterministic mode with no AI model calls — all threshold-based alerting, SNMP polling, flow collection, and synthetic testing continue to function normally.

Provider and Model

FieldDescription
ProviderThe AI provider to use. Currently supported: Anthropic (Claude).
ModelThe specific model identifier (e.g., claude-sonnet-4-20250514). This can be updated as new models are released without requiring a full platform upgrade.

Click Test Connection to verify that WhiteOwl can reach the AI provider's API with the configured credentials. A successful test confirms that your API key is valid and the model is accessible.

Feature Controls

Three independent toggles control which WhiteOwl features use AI:

FeatureDescription
AI AssistantEnables the chat assistant and analysis features. When enabled, users can ask natural language questions about network health, traffic patterns, device status, and more. The assistant dynamically gathers context from PostgreSQL and ClickHouse based on the question content.
Executive ReportsEnables AI-generated summaries in reports. When enabled, report templates like the Executive Summary include narrative analysis, trend interpretation, and actionable recommendations written by the AI model.
Alert IntelligenceEnables AI analysis for alert context. When enabled, AI-powered alert rules (AI Analysis templates) can run holistic network reviews and the alerting engine can provide AI-enriched context for fired alerts.

Each feature can be independently enabled or disabled. For example, you might enable Executive Reports while keeping the AI Assistant disabled if you only want AI for scheduled reporting.


Connection Status

The Connection Status section tests connectivity to WhiteOwl's backend services. Click Test to verify all connections simultaneously.

ServiceDescription
PostgreSQLThe configuration database storing devices, sites, users, alert rules, and credentials. Displays the connection string (e.g., postgres-db).
ClickHouseThe time-series metrics database storing SNMP metrics, flow records, synthetic test results, and log data. Displays the connection URL (e.g., http://clickhouse:8123).
SyntheticsThe synthetic test execution engine. Displays the service URL (e.g., http://192.168.100.132:8091).

Each service shows a status indicator after testing — green for connected, red for unreachable. If a service fails, check that the corresponding container or service is running and that network connectivity is available between the WhiteOwl backend and the service.


API Keys

The API Keys section manages authentication keys for programmatic access to WhiteOwl's MCP (Model Context Protocol) API. The MCP API allows external tools, scripts, and AI systems to query WhiteOwl data.

Creating an API Key

When creating a new API key, you configure:

FieldDescription
Key NameA descriptive label for the key (e.g., "Automation Scripts", "Claude Desktop").
Permission LevelAccess level for the key: Read Only (query data only), Read/Write (query data and execute actions like running tests or acknowledging alerts), or Admin (full access including direct database queries).

Available MCP Tools

API keys grant access to the following tools, with availability based on permission level:

ToolPermissionDescription
get_devicesRead OnlyList all monitored devices with status, vendor, model, and site information.
get_device_detailsRead OnlyGet detailed information for a specific device.
get_interfacesRead OnlyList interfaces for a device with utilization and error statistics.
get_sitesRead OnlyList all sites with device counts.
get_alertsRead OnlyQuery active and historical alerts.
acknowledge_alertRead/WriteAcknowledge an active alert.
top_talkersRead OnlyGet top traffic sources and destinations from flow data.
search_flowsRead OnlySearch flow records with filters.
flow_summaryRead OnlyGet aggregated flow statistics.
get_synthetic_testsRead OnlyList configured synthetic tests.
get_synthetic_resultsRead OnlyGet synthetic test result history.
run_ping_testRead/WriteExecute an on-demand ICMP ping test.
run_tracerouteRead/WriteExecute an on-demand traceroute.
run_http_testRead/WriteExecute an on-demand HTTP test.
get_snmp_metricsRead OnlyQuery SNMP metrics (CPU, memory, interface stats).
get_interface_utilizationRead OnlyGet interface bandwidth utilization data.
network_health_summaryRead OnlyGet an overall network health summary.
query_postgresAdminExecute arbitrary PostgreSQL queries against the configuration database.
query_clickhouseAdminExecute arbitrary ClickHouse queries against the metrics database.
Admin API Keys

Admin-level keys have direct database query access. Only create Admin keys when necessary, and revoke them when no longer needed. For most automation use cases, Read Only or Read/Write permissions are sufficient.

Managing Keys

From the API Keys section you can:

  • Create new keys with a name and permission level.
  • View existing keys, their permission levels, and creation dates.
  • Revoke keys that are no longer needed. Revoked keys immediately lose access.

Claude Desktop

The Claude Desktop section generates configuration for connecting Claude Desktop (Anthropic's desktop AI assistant) to WhiteOwl via MCP. This allows you to chat with Claude on your local machine and have it query live WhiteOwl data — devices, metrics, flows, alerts, and more — through natural language.

Setup

  1. Click Generate Config to produce a JSON configuration block.
  2. Copy the generated configuration.
  3. Paste it into your Claude Desktop MCP configuration file (typically ~/.claude/config.json or via the Claude Desktop settings).
  4. Restart Claude Desktop.

Once connected, Claude Desktop can use all MCP tools available to the configured API key. You can ask questions like "What devices are down?", "Show me top talkers for the last hour", or "Run a ping test to 8.8.8.8" and Claude will query WhiteOwl in real time.


Audit Log

The Audit Log records all MCP API requests for monitoring and troubleshooting. Each log entry includes:

FieldDescription
TimestampWhen the request was made.
API Key NameWhich API key was used for the request.
Tool NameThe MCP tool that was called (e.g., get_devices, search_flows).
StatusWhether the request succeeded or failed: success, error, or warning.
Response TimeHow long the request took to complete (in milliseconds).

The audit log is useful for tracking API usage patterns, identifying slow queries, debugging integration issues, and ensuring API keys are being used appropriately.


API Documentation

The API Documentation section provides ready-to-use code examples for integrating with the MCP API. Examples are available in multiple formats including cURL and Python, with the full tool list and request/response schemas documented.

This section serves as a quick-reference for developers building integrations without needing to leave the WhiteOwl UI.