The process by which an AI agent asks a server what capabilities exist before using them. Unlike static documentation, runtime discovery creates a moving target for security review because new tools can appear without code changes on the client side.
Expanded Definition
Runtime discovery is the capability negotiation step that happens after an AI agent connects but before it invokes a tool, API, or server-side function. In NHI and agentic AI environments, it means the agent does not rely only on a predeclared integration contract; it queries the target system for what is currently available, such as actions, schemas, scopes, or invocation rules. That makes the control plane more flexible, but it also means security review cannot assume a fixed attack surface.
Usage in the industry is still evolving. Some teams treat runtime discovery as a feature of model orchestration, while others treat it as part of API governance or MCP-style tool negotiation. The practical issue is the same: if discovery is not constrained, an agent may see more capability than was approved in design. That is why runtime discovery should be paired with explicit authorization, logging, and allowlisting, consistent with the visibility and governance mindset in the Ultimate Guide to NHIs — Key Challenges and Risks and the access control expectations in NIST Cybersecurity Framework 2.0.
The most common misapplication is assuming a tool is safe because it was not present in the client code, which occurs when server-side capabilities change after deployment without a fresh policy review.
Examples and Use Cases
Implementing runtime discovery rigorously often introduces tighter policy enforcement and more latency at connection time, requiring organisations to weigh adaptability against the cost of additional review and logging.
- An AI support agent queries a ticketing server for available actions and only receives read-only operations during business hours, while elevated actions require separate approval.
- A workflow agent discovers new MCP-style tools after a platform update, so the security team uses the NHI Lifecycle Management Guide to align the new capability with approval, rotation, and revocation processes.
- A developer-facing assistant checks a document service for schema changes at session start, then enforces policy before any write operation is exposed to the agent.
- A security platform monitors discovery responses for unexpected scopes or privileged functions, using the NIST Cybersecurity Framework 2.0 to structure access control and continuous monitoring.
- An operations agent is allowed to discover only a narrow toolset in production, while a broader discovery profile is reserved for test environments to reduce accidental impact.
These use cases are most defensible when paired with the operational lessons in Top 10 NHI Issues, especially where discovery can widen the effective privilege boundary without changing the client.
Why It Matters in NHI Security
Runtime discovery matters because it changes the trust model for every agent that can act on behalf of a system, workflow, or user. If a server can reveal new actions at runtime, then the real security boundary is not just the client binary or prompt template, but the policy that governs what the agent is allowed to learn and invoke. That is especially important for NHIs, where entitlements often outlive design assumptions.
This is where governance breaks down in practice: NHI Mgmt Group reports that only 5.7% of organisations have full visibility into their service accounts. In environments with weak visibility, discovery events can surface new reachable functions long before anyone notices the associated privilege expansion. A discovery response can also expose secrets-handling paths, admin APIs, or write operations that were never intended for routine agent use, which makes NHI Lifecycle Management Guide relevant to approval, review, and revocation discipline.
Organisations typically encounter the operational impact only after a tool update, privilege escalation, or incident review, at which point runtime discovery becomes unavoidable to address.
Related resources from NHI Mgmt Group
Deepen Your Knowledge
Reviewed and updated by the NHIMG editorial team on June 6, 2026.
NHI Mgmt Group — the #1 independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org
