Claude 4 API: Tool Use Streaming & Prompt Caching vs Cursor 1.0

“The primitive here is clean: incremental tool-call deltas over SSE, and a cache-control header you attach to prompt segments to pin them server-side. The DX bet is that complexity lives in the HTTP layer, not in a new SDK abstraction — you opt in per-request, no new mental model required. The moment of truth is calling `stream=true` on a tool-use request and watching partial JSON arguments arrive before the model finishes thinking, which actually matters for agent loops where you want to dispatch work early. This is not a weekend-script replacement — implementing correct incremental JSON parsing for partial tool arguments plus a reliable distributed cache with 2M token capacity is a real engineering problem Anthropic has solved for you. The specific decision that earns the ship: cache invalidation is explicit and cache hits are reflected in the usage object, so you can actually measure what you're saving instead of guessing.”

“The primitive is clear: a persistent agent runtime that survives session close and operates asynchronously against your repo, with team-scoped context as a first-class object — not a settings page. The DX bet is that complexity lives in the agent orchestration layer, not in the developer's config, and mostly that bet pays off. The moment of truth is submitting a background task and closing your laptop; when it's actually done and the diff is clean on return, that's a real product. The specific decision that earns the ship: making team memory a write-path feature, not just retrieval — agents can update shared context, which no weekend Lambda script replicates.”

“Direct competitors are OpenAI's cached completions and Google's context caching in Gemini 1.5 — both shipping for months — so Anthropic is catching up, not leading. The specific scenario where this breaks: cache hit rates depend entirely on prompt structure, and developers who dynamically compose system prompts (inserting user-specific context at the top) will see near-zero cache utilization and pay full price while assuming they're saving money. The prediction: this feature doesn't get killed — it becomes table stakes infrastructure and Anthropic wins by having the largest cache window (2M vs. competitors' current limits). What would have to be true for me to be wrong: OpenAI ships a 10M token cache window before Anthropic's ecosystem matures, commoditizing the advantage. Still a ship because the streaming tool-use delta is genuinely differentiated — no competitor has clean partial-argument streaming for tool calls yet, and that changes agent loop architecture in ways that matter.”

“The direct competitors are GitHub Copilot Workspace and JetBrains AI, both of which are racing toward async agents — Cursor is ahead on shipping something developers can actually demo breaking on a real codebase today. The scenario where this collapses: multi-file refactors across monorepos with conflicting agent tasks, where the shared context model becomes a write-conflict nightmare at 50+ engineers. The 12-month kill condition isn't a competitor — it's GitHub shipping background agents natively into Codespaces with zero additional cost to existing Enterprise customers, which is the most obvious move on their board. What earns the ship anyway: the team context memory is a genuine moat attempt, not just a feature flag on a model API.”

“The thesis this bets on: by 2027, the dominant AI application architecture is a persistent agent with a large, stable context (tools, memory, instructions) that gets reused across thousands of user interactions — making context I/O cost the primary unit economics lever, not generation cost. The dependency that has to hold: agents don't collapse back to stateless chatbots, and context windows keep growing faster than per-token prices fall. The second-order effect nobody's talking about: prompt caching at 2M tokens makes it economically viable to give every enterprise user a fully-loaded, role-specific agent context at request time — which shifts competitive differentiation from 'who has the best model' to 'who has the best cached context corpus,' effectively making knowledge curation the new moat. This tool is riding the trend of context-window expansion-as-infrastructure, and it's on-time, not early — but the streaming tool-use primitive is ahead of the curve on agent loop efficiency. The future state where this is infrastructure: every production agentic system has a cache manifest the same way it has a CDN config.”

“The thesis Cursor is betting on: by 2027, most engineering work is orchestrated asynchronously across human and agent collaborators, and the editor becomes the control plane for that fleet, not just the surface for a single developer's keystrokes. The dependency that has to hold is that context management remains hard enough that a dedicated layer is worth paying for — if model context windows expand to encompass entire large codebases cheaply, the shared memory feature commoditizes. The second-order effect that nobody is talking about: team codebase memory shifts knowledge ownership from senior engineers to the tooling layer, which changes onboarding, attrition risk, and how engineering orgs value individual contributors. Cursor is early on the async multi-agent trend relative to the IDE incumbents, and the infrastructure bet is credible.”

“The buyer is the engineering team at any company running Claude in production with long system prompts or multi-step agents — this comes out of the AI infrastructure budget, not a new budget line, which means no procurement friction. The pricing architecture is sound: cache reads at ~90% discount means the savings are real and measurable in the first billing cycle, which creates immediate retention — developers who restructure prompts to maximize cache hits are now architecturally coupled to Anthropic's caching implementation. The moat question is the honest one: this is infrastructure that OpenAI and Google will match, so the defensible position isn't the feature itself but the ecosystem of developers who've restructured their codebases around it. What survives a 10x model price drop: the streaming tool-use architecture, because that's about latency, not cost. The specific business decision that makes this viable is pricing cache reads as a separate SKU — it lets Anthropic capture value from high-volume production workloads without losing price-sensitive experimenters.”

“The buyer is a VP of Engineering or CTO pulling from a developer tooling or productivity budget — this is not a bottoms-up PLG play anymore, the team collaboration tier signals a deliberate move upmarket. The pricing architecture is sound: individual Pro at $20 creates a personal habit, Business at $40 creates the enterprise conversation, and shared context creates the switching cost because migrating team memory is painful. The moat question is the right one: shared codebase memory creates genuine workflow lock-in if teams actually adopt it, which is a data network effect with teeth. What kills it is if Anthropic or OpenAI decide to bundle a code agent product directly — Cursor's defensibility lives entirely in the editor UX and the memory layer, so they need to compound both faster than model providers commoditize the inference.”