Governance scaled toyour regulatory load.
Every package includes the same core capabilities. Compare request and connector allowances, TG-Sentinel access, and each package's expansion basis.
Evaluation Pilot
- Expansion basis
- Agreed workflow and acceptance criteria
- Governed-request capacity
- 2M included; free usage window up to 10M per 30-day pilot
- Connector allowance
- No fixed selection quota; usage is visibility-only
- TG-Sentinel model access
- Fast + Max
Package features and limits
- Try the core governance capabilities with real traffic
- Create up to 3 Agents and 5 Guards
- Up to two paid 30-day extensions; 90 days maximum total
- No annual payment option
Production Governance
- Expansion basis
- Usage and connector expansion
- Governed-request capacity
- 10M included governed requests per month; overage available
- Connector allowance
- 5 connector selections included; additional connectors available as add-ons
- TG-Sentinel model access
- Fast + Max + Max-Reasoning
Package features and limits
- Production runtime enforcement with rewrite, block, and escalate actions
- Agent DNA and Tool/MCP authorization
- Standard provenance and export
- Standard support
Governance Assurance
- Expansion basis
- Review, retention and enterprise assurance
- Governed-request capacity
- 25M included governed requests per month; overage available
- Connector allowance
- 10 connector selections included; additional connectors available as add-ons
- TG-Sentinel model access
- Fast + Max + Max-Reasoning + Ultra
Package features and limits
- Everything in Production Governance
- Advanced policy and runtime review with human approval
- Compliance reports and provenance verification
- Extended retention and advanced analytics
- Optional dedicated GPU/private-cloud connection
Sovereign Governance
- Expansion basis
- Custom deployment and capacity
- Governed-request capacity
- Contracted/dedicated capacity; no standard request quota
- Connector allowance
- No standard numeric selection limit
- TG-Sentinel model access
- Fast + Ultra + Ultra-ReasoningTG-Sentinel:G1 access when available
Package features and limits
- Everything in Governance Assurance
- Customer-managed independent deployment
- On-premise and air-gapped options
- Custom data residency and key lifecycle
- Custom integrations and contractual SLA
- Confidential Computing Support (Preview)
Sovereign deployment, data residency, key lifecycle, and SLA scope, as well as optional Assurance private-cloud connections, are subject to contract and operational readiness.
Shared across all four packages
Included in every package
- Guard Catalog
- Adaptive evaluation
- Runtime decisions
- Tenant isolation and RBAC
- HMAC provenance
- Customer BYOK model execution
- Cloud HSM (GCP, AWS, Azure)
- Cloud Private Access/Trust-Zone Support
- Agent DNA
- Basic analytics
Feature-by-featureAI governance comparison.
Compare documented capabilities across each platform. Detailed Theseus Guard configuration and availability are confirmed in a sales consultation.
| Feature | Microsoft Purview + Agent 365 | IBM watsonx.governance | OneTrust AI Governance | Holistic AI | Credo AI | Theseus Guard(Stateful) |
|---|---|---|---|---|---|---|
| AI / agent inventory and identity | Yes | Partial | Yes | Yes | Yes | Yes (Agent DNA) |
| Policy and risk workflows | Yes | Yes | Yes | Yes | Yes | Yes |
| Runtime policy enforcement | Partial | Partial* | Yes | Yes | Preview | Yes |
| Agent / tool action controls | Partial | Yes* | Partial | Partial | Preview | Yes (Tool/MCP) |
| Skill-scope security controls | Not stated (1) | Partial (2) | Not stated | Not stated | Not stated (3) | Yes (skill-scope policy) |
| Cross-turn / session risk | Not stated | Not stated | Not stated | Not stated | Not stated | Yes |
| Human review / approval workflow | Partial | Not stated | Yes | Not stated | Not stated | Yes |
| Audit trails / compliance evidence | Yes | Yes | Yes | Yes | Partial | Yes |
| Tamper-evident provenance | Not stated | Not stated | Not stated | Not stated | Not stated | Yes (HMAC) |
| Decision evidence linked to runtime verdict | Partial | Partial | Partial | Partial | Preview | Yes |
Competitor comparison last reviewed: September 23, 2026. Claims reflect official public materials reviewed by this date; vendor products and documentation may change.
More from Theseus Guard
Additional Theseus Guard capabilities
Zero-Data AI Data Plane (ZT-ADP)
Tenant-admin configurable Disabled, Observe, and Enforce modes with Format-Preserving Encryption across prompt, tool, MCP, and skill scopes for 100+ sensitive entity classes.
Tenant identity and isolation
JWT and API-key access with role-based controls, tenant-scoped requests and data, and database, cache, queue, and provenance isolation.
Governance Guard lifecycle
Compose Kernel, Identity, and Probe Sets; validate compatibility; manage versioned releases, impact previews, and tenant sync while preserving tenant overrides.
Enterprise and department governance
Company-wide Enterprise Kernels and department-specific Guards, with SCIM/directory sync, employee and department authorization, access grants, and identity history.
Agent lifecycle and model controls
Agent DNA profiles and deployment approvals, scoped agent keys, provider/model policies, controlled fallback, and provider-spend reconciliation.
OpenAI-compatible governance gateway
Drop-in chat-completions and models endpoints with pre-broker tool authorization, policy enforcement, model routing, and decision provenance.
Adaptive evaluation and monitoring
L1 Precedent Decision Cache, L2/L3 evaluation with TG-Sentinel models (including G1 MoE), sector probes and confidence benchmarks, drift monitoring, and anomaly alerts.
Policy operations and remediation
Shadow-to-promotion policy workflows, runtime reviews, and Advanced Review & Execution with approval, authorization, execution, verification, rollback, and escalation.
Governance analytics and reporting
Governance Health Index, posture dashboards, alert assignment and SLA tracking, plus compliance-report generation, comparison, and export.
Privacy and sensitive-data controls
PII detection, masking, tokenization, and audited disclosure. The token vault is an MVP control, not durable enterprise key management.
Integrations and adapters
ServiceNow, OneTrust, Elastic SIEM, signed webhooks, LangChain/DeepAgents, and plan-aware billing. External credentials and compatible versions are required; live Stripe and Azure Key Vault/HSM/Monitor are deployment-dependent.
Collaboration and workflow integrations
Connect Microsoft Teams, Slack, and Jira with Theseus Guard to fit your AI governance workflows. TG-Agent (Private Beta) gives selected users non-production access to tenant-admin-approved Agent DNA agents through its Web UI or CLI. Access is scoped by department and user, and supported coding-agent clients can use compatible model names or aliases.
Deployment options
Private-network shared serving, optional dedicated private-cloud deployment, and customer-managed on-premise or air-gapped Sovereign deployment; availability depends on package and deployment configuration.
Confidential Computing Support
Available with Sovereign Governance (Preview); scope and availability are confirmed in a sales consultation.
Runtime policy simulator
Preset-scenario governance testing with exportable JSON results; this is a non-production simulator, not a production runtime service.
Yes = stated in product materials. Partial = related, narrower, or integration-dependent. Preview = research preview. Not stated = not found in the linked competitor sources; this does not prove the capability is unavailable. Competitor cells use linked official vendor sources; Theseus Guard cells follow current product materials.
* IBM agent/tool runtime controls are documented for watsonx Orchestrate, an adjacent product to watsonx.governance.
(2) IBM's adjacent watsonx Orchestrate ADK validates uploaded Python scripts used by skills and runs them in isolation; this is not a watsonx.governance feature. See the official skill management documentation.
(1) Microsoft's Agent 365 Skills are guided setup workflows; reviewed documentation does not describe security controls over installed skill packages.
(3) Credo's Agent Governor paper describes generated skills as advisory artifacts; reviewed documentation does not describe controls over arbitrary installed skill packages.
Third-party product names and trademarks belong to their respective owners and are used solely for identification. No affiliation or endorsement is implied.