IBM announced new controls for AI agents at TechXchange on October 27. Agent Identity gives each agent a record of its owner, status and authority. HashiCorp Vault 2.2 adds secured agent accounts and credential controls. Both are generally available.
IBM says Agent Identity can share identity and authorization details with Vault. Vault can then issue short-lived credentials that limit an agent to the access it needs. If an agent tries to act outside its approved permissions, access can be denied and its session revoked in real time. Vault 2.2 also automatically renews agent credentials and incorporates post-quantum encryption.
Orchestrate adds discovery and testing
IBM also announced tools for managing agents in watsonx Orchestrate. A discovery feature, now in tech preview, connects to platforms including Amazon Bedrock, Microsoft Foundry and Google Gemini to find agents built there. It also covers models and MCP servers.
A containment feature in tech preview runs agents in separate sandboxes and gives teams a way to shut down an agent. Evaluation capabilities are generally available. IBM says teams can test agents, models and MCP servers before deployment, monitor them in production, and track each agent’s model and token use.
Orchestrate works with Agent Identity and Vault to set an agent’s authority, control its credentials and audit its actions. IBM says the combined controls can revoke an agent’s trust and credentials when risk changes.
Other announced tools include Secure Hub, a public preview for viewing operational and security information across Vault environments, and IBM Integration Intelligence Hub, a tech preview that gives agents context about existing APIs and systems. Secure Hub is in public preview, while Integration Intelligence Hub and Orchestrate’s discovery and containment features are in tech preview.
