Rebeca Moen
Apr 10, 2026 19:10
Anthropic engineers element how they construct and refine AI agent instruments for Claude Code, introducing progressive disclosure strategies that form AI improvement.
Anthropic has pulled again the curtain on how its engineering staff designs instruments for Claude Code, the corporate’s AI-powered software program improvement assistant. The detailed technical breakdown, revealed April 10, gives uncommon perception into the iterative course of behind constructing efficient AI agent techniques.
The $380 billion AI security firm’s strategy facilities on what engineer Thariq Shihipar calls “seeing like an agent” — primarily understanding how an AI mannequin perceives and interacts with the instruments it is given.
Trial and Error with AskUserQuestion
Constructing Claude’s question-asking functionality took three makes an attempt. The staff first tried including a query parameter to an current device, which confused the mannequin when person solutions conflicted with generated plans. A second try utilizing modified markdown formatting proved unreliable — Claude would “append further sentences, drop choices, or abandon the construction altogether.”
The successful answer: a devoted AskUserQuestion device that triggers a modal interface, blocking the agent’s loop till customers reply. The structured strategy labored as a result of, as Shihipar notes, “even the perfect designed device would not work if Claude would not perceive how one can name it.”
When Instruments Develop into Constraints
The staff’s expertise with activity administration reveals how mannequin enhancements can render current instruments out of date. Early variations of Claude Code used a TodoWrite device with system reminders each 5 turns to maintain the mannequin on observe.
As fashions improved, this grew to become counterproductive. Claude began treating the todo checklist as immutable somewhat than adapting when circumstances modified. The answer was changing TodoWrite with a extra versatile Process device that helps dependencies and cross-subagent communication.
From RAG to Self-Directed Search
Maybe essentially the most important shift concerned how Claude finds context. The preliminary launch used retrieval-augmented era (RAG), pre-indexing codebases and feeding related snippets to Claude. Whereas quick, this strategy was fragile and meant Claude was “given this context as a substitute of discovering the context itself.”
Giving Claude a Grep device modified the dynamic completely. Mixed with Agent Expertise — which permit recursive file discovery — the mannequin went from being unable to construct its personal context to performing “nested search throughout a number of layers of recordsdata to search out the precise context it wanted.”
The 20-Instrument Ceiling
Claude Code at the moment operates with roughly 20 instruments, and Anthropic maintains a excessive bar for additions. Every new device represents one other choice level for the mannequin to guage.
When customers wanted Claude to reply questions on Claude Code itself, the staff averted including one other device. As a substitute, they constructed a specialised subagent that searches documentation in its personal context and returns solely the reply, retaining the primary agent’s context clear.
This “progressive disclosure” strategy — letting brokers incrementally uncover related data — has turn into central to Anthropic’s design philosophy. It echoes the corporate’s broader deal with creating AI techniques which might be useful with out turning into unwieldy or unpredictable.
For builders constructing their very own agent techniques, the takeaway is obvious: device design requires fixed iteration as mannequin capabilities evolve. What helps an AI right now may constrain it tomorrow.
Picture supply: Shutterstock