How Postman Scales Agent Mode on Amazon Bedrock: Tools, Context and Controls

On October 9, 2026, Postman and AWS shared architectural lessons from running Agent Mode in a mature API-development platform. Postman serves a community of 40 million developers; that figure is not a claim that all of them use Agent Mode.

The central challenge was not simply model quality. Early designs exposed many small tools, forcing long chains of calls. Postman reported more tool-selection errors once the visible set exceeded roughly 40, including incorrect arguments and nonexistent tool calls.

TECHNICAL CONTEXT: The announced approach needs to be understood in its specific technical and operational context. A useful evaluation begins by identifying the exact task, the information available to the system and the expected outcome.

IMPLEMENTATION CONSIDERATIONS: The practical value depends on how the system is integrated with existing processes and controls. Teams should identify which actions are permitted, how failures are detected and who can review consequential results.

EVALUATION AND LIMITS: The stated capabilities and figures should be evaluated under their reported conditions. Independent tests and representative real-world tasks help establish whether the approach is suitable beyond a demonstration.

PRACTICAL EVALUATION: Before adopting this technology, teams should define a specific workflow and measurable success criteria. A limited pilot can compare completion time, output quality and recovery from failures against the existing process. A successful demonstration is only one step toward a dependable deployment.

SECURITY AND OPERATIONS: Systems involving AI or automation require attention to source accuracy, user permissions, audit trails and ways to stop or reverse actions. Workflows affecting external services or production infrastructure need stronger controls than a local prototype. Operational responsibility remains with the deploying organization.

ANNOUNCEMENT VERSUS AVAILABILITY: Claims in a product announcement depend on the stated conditions, test environment and release stage. Preview features and experimental findings should not be presented as broadly available production results. Readers should verify current limitations and eligibility in the primary source.

WHAT TO WATCH: The long-term value depends on integration with existing work, cost, reliability and the ability to verify results. Organizations should track real deployments and repeat evaluations as products change, rather than rely solely on initial demonstrations.

Its architecture now narrows more than 170 candidate tools to about 15 relevant ones per task. It also reduces dependencies on open tabs and interface state so an agent can work with application data rather than imitate a human navigating screens.

Context quality proved equally important. Instead of passing UI-oriented data structures directly to the model, Postman built handlers that distill the information needed for each task. Amazon Bedrock supplies model flexibility, geographic inference controls and prompt caching.

State-changing actions require user approval, and Bedrock Guardrails can redact personal information. The case shows that dependable agents depend on permissions, context engineering and interaction design—not model choice alone.