An agent runtime deployed on infrastructure controlled by the organisation rather than the vendor. This gives teams more flexibility and model choice, but it also makes them responsible for access control, secrets handling, monitoring, and secure operations across the full lifecycle.
Expanded Definition
A self-hosted agent is an AI or automation runtime operated on infrastructure the organisation controls, such as its own cloud tenancy, private cluster, or on-prem environment. In NHI practice, that distinction matters because the organisation inherits responsibility for the agent’s identity, tool permissions, runtime hardening, and secret lifecycle rather than depending on a vendor-managed boundary.
Definitions vary across vendors, especially when an “agent” is really a hosted API wrapper with limited execution authority. In NHI Management Group’s view, the term should be reserved for runtimes that can independently act, call tools, and retain credentials under organisational governance. That makes it closely related to the control expectations discussed in the OWASP NHI Top 10 and the OWASP Agentic AI Top 10, where execution authority and tool access are central risk drivers. The most common misapplication is calling any internally deployed chatbot “self-hosted” when it actually relies on vendor-managed identity, logging, or secret storage.
Examples and Use Cases
Implementing a self-hosted agent rigorously often introduces operational overhead, requiring teams to weigh autonomy and data control against patching, monitoring, and identity governance costs.
- A customer-support agent runs in a private Kubernetes cluster, using short-lived tokens to read case data and create tickets while the organisation enforces access review and rotation policies.
- An engineering agent executes code-review tasks inside an internal CI/CD environment, but only after receiving scoped credentials from a managed secrets platform and passing policy checks.
- A finance reconciliation agent is hosted in the company’s own cloud account so transaction data never leaves the tenant, yet the team must still log every tool call and restrict lateral movement.
- A security operations agent uses internal retrieval systems to correlate alerts, with identity federation and least privilege aligned to guidance from the NIST AI Risk Management Framework.
- After incidents such as the CoPhish OAuth Token Theft via Copilot Studio, teams often move agent workloads in-house to tighten control over credentials and runtime policy.
Why It Matters in NHI Security
Self-hosting does not reduce NHI risk by itself; it shifts accountability inward. The organisation now owns the full attack surface: service-account design, secrets exposure, tool permissions, logging, model routing, and offboarding. That matters because NHIs are frequently overprivileged and under-observed. NHI Management Group research shows that 97% of NHIs carry excessive privileges, and 79% of organisations have experienced secrets leaks, with 77% resulting in tangible damage. A self-hosted agent that can reach production systems with a long-lived key becomes an internal breach path if lifecycle controls are weak.
This is where zero trust and agent governance converge. The operator must assume the agent can be prompt-injected, misrouted, or induced to overuse credentials, then constrain it with segmentation, ephemeral identity, and auditability. The same concern appears in incident analysis such as the Moltbook AI agent keys breach and the NIST AI Risk Management Framework, both of which reinforce that operational control must be matched by governance control. Organisations typically encounter the need for strict self-hosted agent governance only after a credential leak, unsafe tool action, or unexpected data access, at which point the term becomes operationally unavoidable to address.
Standards & Framework Alignment
This section maps relevant standards and security frameworks to the operational risks and controls described in this guidance.
OWASP Non-Human Identity Top 10, OWASP Agentic AI Top 10 and CSA MAESTRO address the attack and risk surface, while NIST AI RMF and NIST Zero Trust (SP 800-207) set the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| OWASP Non-Human Identity Top 10 | NHI-02 | Self-hosted agents expand secret-handling and lifecycle risks under NHI controls. |
| OWASP Agentic AI Top 10 | A2 | Agent tool access and execution authority are core concerns for self-hosted runtimes. |
| NIST AI RMF | NIST AIRMF frames governance, measurement, and risk treatment for AI systems. | |
| NIST Zero Trust (SP 800-207) | PE-3 | Self-hosted agents fit zero trust assumptions about continuous verification and segmentation. |
| CSA MAESTRO | MAESTRO addresses agentic AI threat modeling and runtime governance needs. |
Inventory the agent’s credentials, rotate them, and enforce least privilege across every tool it can call.
Related resources from NHI Mgmt Group
Deepen Your Knowledge
Reviewed and updated by the NHIMG editorial team on August 19, 2026.
NHI Mgmt Group — the #1 independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org
