Platform Platform
System
ConceptsEnginePolicy as codeDeclarationsSafe changeGatewaysIntegrationsObservabilityAdministrationSecurityHuman reviewAudit and evidenceData retentionSecrets and data classification
Controls
Registries and documentationAuthentication and authorizationInjection detectionData redactionCode fingerprintingRole and judge checksContent classificationSpend and loop limitsBusiness rules
Solutions Solutions
By what you do
Sell into the enterpriseControl the AI you run
By industry
Financial servicesDigital assetsInsuranceHealthcareLegalUser-generated content
By discipline
AI governanceTrust and safetyRisk and compliance
Cases Cases Embedded control planeSource-code leakTrading agents over MCPLive firehoseRefund assistant
Compare Compare LiteLLMNVIDIA NeMo GuardrailsOPAROOSTAgent Governance Toolkit
Resources Resources
Guides
Enterprise review questionsPrompt injectionAgent and control layerAgent architecturesDecision system mapAI control maturity model
Standards
Standards OWASP Agent Control StandardEU AI ActPMI AI standardNIST AI RMFERC-8004
Book a demo
Compare

Microsoft gives it away free. Here is what you still build and run.

The Agent Governance Toolkit is an open-source project worth reading to understand the problem. It is a public-preview toolkit you wire into your own agents, so state, policy lifecycle, and operations are yours to build and run. Swiftward is the finished on-prem product.

What the toolkit is

A free, MIT-licensed set of framework adapters and SDKs with runtime enforcement. Microsoft building it confirms this is a real category.

What Swiftward does differently

  • You integrate the toolkit into your agent code and operate it. Swiftward is a single binary plus Postgres with the console, gateways, and audit already in place.
  • The toolkit's kernel is stateless by design, so you bring and run your own state store (Redis, Kafka, and the like) for counters, rate limits, and trust scores. Swiftward ships that state in Postgres, ACID and safe to re-process.
  • Versioned policy-as-code is the part we add: immutable versions, rollback by naming the previous version, shadow mode and backtesting against real traffic, and the record of a past decision on the exact version that was live.
  • Human review here survives a restart: the case waits, and the decision on it re-enters the same audited pipeline as a recorded event.

How far it reaches

You wire the toolkit into agent code, so it governs the agents your own teams write. Swiftward sits in the path of the call, so one policy also covers an agent you bought inside a product whose code you cannot read, and a payment or a claim decided by ordinary application code with no agent in it at all. See the capability comparison.

Based on the Microsoft Agent Governance Toolkit's public documentation as of April 2026. It moves fast; if something here has changed since, tell us and we will update.
Book a demo