Part 30: Identity and governance for agents across five vendors
Part 30 · One Agent, Five Platforms Five platforms means five policy models, unless you design one. Each vendor has its own idea of who an agent is, where its permissions live, what gets logged and where the agent list is kept. If you run agents on more than one, governance has to sit above them. This part puts the five builds side by side on six controls, and then maps one policy model onto all…
Part 30 explores the challenges of managing agents across five different vendor platforms. Each vendor has its own approach to defining an agent, managing permissions, logging, and tracking agents. Governance must sit above these differences to ensure consistent policies.
The builds examined six key controls: agent identity, least privilege, audit trails, PII/data residency, a central registry of agents and tools, and a unified policy model. These controls were mapped to the five vendor platforms.
For agent identity, Azure assigns new Foundry agents their own Entra agent identity with no initial permissions, while AWS and SAP utilize different methods. MuleSoft and Boomi have varying approaches, with Boomi lacking a distinct agent identity altogether.
Least privilege is achieved by enforcing tool-specific permissions for each agent. This is enforced at the gateway level where possible, and in the process API regardless of vendor settings. The approval_id check remains in place on every platform.
Audit trails are crucial for compliance, but they vary across platforms. MuleSoft offers OpenTelemetry traces, while Azure and AWS utilize OpenTelemetry telemetry in different logging systems. SAP provides gateway logs and traces.
PII and data residency also differ. MuleSoft has an LLM PII detection policy, while SAP lists data masking and content filtering. Boomi offers word filters and custom regex rules but lacks a dedicated PII feature. Azure and AWS were not checked for PII features during the builds.
Each vendor maintains its own agent and tool registry, with varying degrees of coverage. The policy model recommends using a single registry of record, where all agent and tool information is centralized. This registry should encompass ownership, identities, tools, scopes, data classes, regions, and associated agent cards.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.