Progressive Discovery

Expanded Definition

Progressive discovery is a capability-loading pattern used in MCP and agentic systems to expose tools, scopes, or actions only when a task actually requires them. Rather than presenting an agent with a full tool catalog up front, the system narrows what is available at each step, which reduces context bloat and helps keep the effective permission set smaller.

In NHI governance, this pattern is useful because it changes how authorization is staged: the agent may know a capability exists, but it does not receive the full operational surface until policy, intent, and task state justify it. That makes progressive discovery closely related to NIST SP 800-53 Rev 5 Security and Privacy Controls concepts such as least privilege and access enforcement, even though no single standard governs this pattern yet. Usage in the industry is still evolving, especially where tool brokers, agent runtimes, and policy engines split responsibility for what is discovered, approved, and executed.

The most common misapplication is treating progressive discovery as a user-interface choice, which occurs when teams hide tools visually but leave broad machine access unchanged.

Examples and Use Cases

Implementing progressive discovery rigorously often introduces orchestration complexity, requiring organisations to weigh tighter control of agent reach against additional policy logic and runtime checks.

• An internal coding agent first receives only read-only repository tools, then gains deployment actions only after a change request is approved and the task moves into release preparation.
• A procurement agent discovers supplier lookup and contract review tools before payment tools, limiting exposure while it validates the request using the Ultimate Guide to NHIs — Key Challenges and Risks guidance on blast-radius reduction.
• A support copilot loads ticketing, knowledge base, and identity verification tools in sequence, so each tool becomes available only after the preceding step confirms the request context.
• An agent platform uses policy-driven discovery to reveal secrets-handling actions only to a bounded workflow, aligning the runtime with the lifecycle discipline described in the NHI Lifecycle Management Guide.
• A security operations agent surfaces containment and enrichment tools only after alert severity crosses a threshold, which keeps the default action set narrow and auditable.

These patterns are most effective when discovery is coupled to task state, identity assurance, and explicit policy decisions rather than simple role membership.

Why It Matters in NHI Security

Progressive discovery matters because overly broad tool exposure is one of the easiest ways for an agent to exceed intent, whether through prompt injection, workflow confusion, or accidental execution. When capabilities are loaded only as needed, the organisation reduces the chance that a compromised or misrouted agent can touch high-risk actions such as secret retrieval, deployment, or privilege escalation. That is especially important in environments where NHIs already suffer from over-permissioning and poor visibility, as highlighted in NHIMG research on NHI risk patterns and the Top 10 NHI Issues.

This is not just a design preference. NHI Mgmt Group reports that 97% of NHIs carry excessive privileges, which makes reducing the default surface area a practical control objective rather than a theoretical one. Progressive discovery supports stronger segmentation, narrower audit trails, and more defensible just-in-time access decisions, especially when paired with NIST SP 800-53 Rev 5 Security and Privacy Controls for access enforcement and monitoring.

Organisations typically encounter the impact only after an agent unexpectedly reaches a sensitive tool or data source, at which point progressive discovery becomes operationally unavoidable to address.

Related resources from NHI Mgmt Group

• Why is NHI discovery and inventory the primary goal of NHI security?
• Why is continuous discovery of AI agents important?
• When does secrets discovery become insufficient on its own?
• When should organizations consider adopting advanced tool discovery for AI agents?