Agents that call tools to finish multi-step tasks: the ReAct loop as a durable workflow, JSON-schema tool calls, microVM sandboxes, per-user OAuth, approvals, injection defences and pass^k evals: one-hour boards at three levels.
Engineers designing AI products and the systems that serve them.
Your approach: Explain the AI behavior and system boundaries in the brief, with evidence for the requested quality, privacy, latency and cost.
Design an AI Agent Platform with Tool Calling. Agents that call tools to finish multi-step tasks: the ReAct loop as a durable workflow, JSON-schema tool calls, microVM sandboxes, per-user OAuth, approvals, injection defences and pass^k evals: one-hour boards at three levels. Work from the scoping questions below. State assumptions for any unspecified load, guarantee or target, then trace your design end to end. Explain one difficult case and a credible alternative; the worked example is a reference, not a required implementation.
Resolve the scoping questions for an AI Agent Platform with Tool Calling. Separate stated behavior from assumptions, and identify what is outside your design.
Declare relevant volume, latency, freshness, quality or cost targets with units. Show calculations or an evaluation plan that can test them; unspecified targets are your assumptions, not hidden pass criteria.
Explain how your guarantees hold in a difficult case relevant to this subject. Address: Remember users? Our own model?
Identify users, required behavior and exclusions. Answer: Who uses it? Acting as whom?
Define the information owned by the system and the inputs, outputs and errors at its boundaries. Resolve: Which tools? How long do runs last?
Estimate the dominant workload and resource demand with units and explicit assumptions. For a learned system, also state how quality is measured and what data is available.
Draw or describe the responsibilities needed for an AI Agent Platform with Tool Calling. Trace a representative request, event or job from its input to a visible result; identify durable state owners.
Walk through a difficult case step by step, including detection and recovery. Consider: Watch live? Limits? Remember users? Our own model?
Compare a credible alternative using your chosen workload and guarantees. Explain a remaining risk, a signal to watch and when you would change the design.