IBM announced agent identity, access controls and management features at TechXchange, with some capabilities generally available and others entering preview.
IBM announced a package of enterprise AI-agent management and security capabilities on October 27, 2026, at its TechXchange conference in Atlanta. The tools are intended to help organizations identify agents, set boundaries on their access and monitor their work across business systems.
HashiCorp Vault 2.2 and IBM Agent Identity are generally available. IBM says the combination can give each agent a record of its owner, status and granted authority, while Vault issues short-lived credentials with limited permissions. Access can be denied and a session revoked if an agent attempts an action outside its authorization. Vault 2.2 also adds secured agent accounts, automatic credential renewal and post-quantum encryption.
IBM is also extending watsonx Orchestrate beyond agents built within its own environment. Its discovery and containment functions are in tech preview: discovery can locate agents, models and MCP servers across services including Amazon Bedrock, Microsoft Foundry and Google Gemini, while containment runs agents in isolated sandboxes and provides a shutdown control. Orchestrate’s evaluation capabilities are generally available and can test and monitor agent quality, accuracy and compliance, with cost tracking for model and token use.
The announcement addresses a practical governance problem for companies deploying agents across multiple cloud and software platforms: identity, permissions and monitoring may otherwise be spread across separate systems. IBM’s release combines credential controls with agent discovery and evaluation, rather than requiring every agent to be built in one environment. The availability differs by feature, so the cross-platform discovery and containment functions are not yet generally available. IBM says the controls are designed to help businesses track agent ownership and actions and withdraw access when risk changes.