Skip to main content
AI Agent Catalog

Seven Autonomous AI Agents

The definitive reference catalog for every Contivra AI Agent. Understand what each agent does, what it consumes and what it produces.

Discovery Agent

Always-on asset discovery across your entire data estate

Purpose

The Discovery Agent continuously crawls every connected data source to identify new assets, detect schema changes and surface previously unknown datasets. It operates on configurable schedules and triggers, ensuring the Enterprise Context Layer is always current—without requiring manual cataloging effort from any team.

Responsibilities

  • Crawl all connected data sources on configurable schedules and event triggers
  • Detect net-new assets and register them in the Enterprise Context Layer
  • Identify schema changes and surface change events to downstream agents
  • Generate initial classification and domain tagging suggestions
  • Monitor connector health and alert on connectivity failures
  • Maintain discovery lineage—recording when and how each asset was found

Inputs

Connector configurationsSource schemasDiscovery schedulesExclusion policiesChange detection thresholds

Outputs

Asset recordsSchema change eventsClassification suggestionsDiscovery health reportsLineage records

Key Capabilities

Supports 50+ enterprise connector types including cloud warehouses, databases, data lakes and streaming platforms
Schema inference with data type detection and relationship identification
Change detection with configurable sensitivity thresholds
Discovery lineage tracking—full audit of when every asset was first discovered

Classification Agent

Automated sensitivity, PII detection and domain tagging at enterprise scale

Purpose

The Classification Agent applies PII detection, sensitivity labelling and domain classification to every asset discovered by the Discovery Agent. It uses a combination of pattern matching, policy rules and LLM reasoning to classify at column level—and routes exceptions requiring human judgment to the steward queue.

Responsibilities

  • Detect PII and sensitive data patterns at column level across every discovered asset
  • Apply sensitivity labels (Public, Internal, Confidential, Restricted) based on policy definitions
  • Assign domain classification using semantic taxonomy and LLM inference
  • Re-classify assets automatically when policy definitions are updated
  • Route classification exceptions and low-confidence results to steward review queues
  • Maintain classification history and version audit trails

Inputs

Raw asset metadataColumn samples (anonymised)Policy definitionsDomain taxonomiesLLM classification prompts

Outputs

Classification tagsSensitivity labelsDomain assignmentsConfidence scoresSteward exception queue items

Key Capabilities

Column-level PII detection for 30+ PII categories including GDPR and HIPAA-defined data types
Multi-label classification—an asset can belong to multiple domains and sensitivity tiers
LLM-assisted classification for ambiguous cases with traceable reasoning
Policy-driven reclassification—changes to policies trigger automatic reclassification runs

Documentation Agent

Business-friendly asset descriptions and usage guidance, generated at scale

Purpose

The Documentation Agent eliminates the documentation debt that accumulates in every enterprise data estate. It generates plain-language descriptions, usage guidance and example queries for every asset—drawing on schema, sample data and domain context from the knowledge graph to produce documentation that is accurate, relevant and immediately useful.

Responsibilities

  • Generate plain-language descriptions for every table, view, column and dataset
  • Create persona-appropriate usage guidance for analysts, scientists and architects
  • Suggest example queries and data access patterns based on schema and usage history
  • Enrich assets with domain context and related concept references from the knowledge graph
  • Flag assets with insufficient or low-quality documentation for steward attention
  • Regenerate documentation automatically when schemas change

Inputs

Schema definitionsAnonymised sample dataDomain context from knowledge graphExisting descriptionsUsage query history

Outputs

Asset descriptionsColumn-level documentationUsage notesExample queriesDocumentation quality scoresSteward attention flags

Key Capabilities

Multilingual documentation generation for global enterprise teams
Persona-aware content—documentation tuned for technical vs business audiences
Context-enriched descriptions drawing on the full enterprise knowledge graph
Quality scoring with automatic flagging of incomplete or inconsistent documentation

Governance Agent

Real-time policy compliance monitoring, enforcement and violation alerting

Purpose

The Governance Agent makes governance a continuous, automated process rather than a periodic manual audit. It monitors every data access event against defined policies, flags violations in real time and produces the audit trails that regulators, compliance teams and external auditors require.

Responsibilities

  • Monitor every data access event against defined access and classification policies in real time
  • Detect and alert on policy violations with severity scoring and escalation paths
  • Enforce data retention, classification and quality policies automatically
  • Generate GDPR, HIPAA, SOX and CCPA compliance audit trails
  • Propose remediation steps for flagged violations with priority guidance
  • Notify asset owners and stewards when governance events require attention

Inputs

Policy definitionsAccess logs and eventsAsset metadata and classificationsLineage graphRegulatory rule sets

Outputs

Compliance status reportsViolation alerts with severity scoresImmutable audit trailsRemediation suggestionsRegulatory compliance summaries

Key Capabilities

Real-time event monitoring with sub-second violation detection
Multi-regulation support—GDPR, HIPAA, SOX, CCPA and custom regulatory frameworks
Immutable audit log with cryptographic integrity verification
Automated remediation for routine violations with configurable intervention thresholds

Quality Agent

Continuous data quality monitoring across five dimensions

Purpose

The Quality Agent profiles datasets, scores quality across five dimensions and monitors for statistical drift—ensuring that every data product consumer always knows the trustworthiness of what they are consuming. It surfaces quality issues before they impact downstream AI models, dashboards or business decisions.

Responsibilities

  • Profile datasets for completeness, accuracy, freshness, uniqueness and validity
  • Score quality against defined SLAs and flag breaches with owner notification
  • Establish statistical baselines and monitor for data drift against them
  • Surface quality trends and identify assets at risk of SLA breach
  • Alert data product owners when quality scores approach breach thresholds
  • Generate quality improvement recommendations with prioritised action lists

Inputs

Dataset samplesQuality rules and SLA definitionsHistorical quality baselinesData product contracts

Outputs

Quality dimension scoresSLA compliance statusDrift alertsQuality trend reportsOwner notificationsImprovement recommendations

Key Capabilities

Five-dimension quality scoring: completeness, accuracy, freshness, uniqueness, validity
Statistical drift detection using configurable baseline windows and sensitivity
SLA-aware monitoring—alerts triggered before breach, not after
Quality lineage—tracks how quality scores change through transformation pipelines

Build Agent

Automated data product assembly, documentation and publication

Purpose

The Build Agent reduces the engineering effort required to publish governed, certified data products. Given a product specification, it assembles the product from governed source assets, validates quality requirements, generates documentation and registers the product in the Data Product Registry—following defined contracts and SLAs throughout.

Responsibilities

  • Assemble data products from governed source assets following product specifications
  • Validate quality requirements and SLA thresholds before publication
  • Generate product documentation, schema definitions and data contracts
  • Register published products in the Data Product Registry with version metadata
  • Monitor published products against their contracted SLAs continuously
  • Notify product owners and subscribers of version changes and quality events

Inputs

Product specificationsGoverned source assetsGovernance policiesQuality SLA definitionsContract templates

Outputs

Published data productsProduct contractsVersion metadataDocumentationQuality validation reportsRegistry entries

Key Capabilities

Automated contract generation from product specification templates
Pre-publication quality gate validation against defined SLA thresholds
Semantic versioning with consumer notification on breaking changes
Continuous post-publication SLA monitoring with owner alerting

Impact Agent

Downstream blast-radius analysis before any change is applied

Purpose

The Impact Agent analyses the downstream blast radius of any proposed schema change, policy update or classification revision before it is applied. By traversing the full lineage graph, it identifies every affected asset, data product, AI model and dashboard—enabling teams to make informed decisions about changes that could otherwise cause silent, cascading failures.

Responsibilities

  • Traverse the full lineage graph to identify all downstream dependants of a proposed change
  • Score impact severity based on asset type, consumer criticality and dependency depth
  • Generate a prioritised impact report with affected asset inventory
  • Alert owners of affected assets, data products and AI models
  • Suggest mitigation paths for high-severity impacts
  • Maintain a history of impact assessments for governance and audit purposes

Inputs

Proposed schema or policy changeFull lineage graphAsset metadataConsumer subscription recordsCriticality classifications

Outputs

Impact reportAffected asset inventorySeverity-scored risk assessmentOwner notificationsMitigation suggestionsAssessment audit record

Key Capabilities

Full graph traversal across unlimited lineage depth
Severity scoring weighted by asset type, consumer count and business criticality
What-if analysis—assess multiple change scenarios before committing to any
Integration with change management workflows for pre-change approval gates

Get Started

Ready to Build Trusted
Enterprise AI?

See how Contivra transforms your fragmented enterprise data into a foundation for AI you can actually trust.