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 here is clear: a persistent agent process that can hold context across a multi-step task and write code to disk without you babysitting it — that's a meaningfully different thing from a tab-complete suggestion. The DX bet Cursor made is to own the editor layer entirely rather than be a plugin, which means they control the full context window: open files, terminal state, git diff, the whole workspace. That bet is paying off because the Background Agent doesn't have to serialize state through a plugin API; it just has it. First-10-minutes test: you can open a repo, describe a feature, and watch it work while you review something else — that's not a demo, that's a workflow shift. The specific decision that earns the ship is building the agent runtime inside the editor process rather than as a sidecar service; that's the right architecture and most competitors haven't figured it out yet.”

“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.”

“Direct competitor is GitHub Copilot Workspace, and Cursor's Background Agent beats it on one specific dimension: the agent operates inside your actual editor state rather than a sandboxed PR branch with limited context. The scenario where this breaks is large monorepos with complex build systems — the agent loses coherence when the dependency graph is deep and the feedback loop from running tests takes more than a few seconds. What kills it in 12 months isn't a competitor; it's that Anthropic and OpenAI are both building coding agents that don't require you to be inside a specific editor. Cursor's moat is the editor context, and that moat holds only as long as VS Code-compatible editors remain the dominant dev environment. For now, the moat is real, the product is genuinely differentiated, and the enterprise audit-log feature is the kind of thing that unblocks procurement — that earns a ship.”

“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 1.0 is betting on: within 3 years, the primary unit of developer work shifts from 'writing code' to 'reviewing and directing code,' and the editor that owns that review surface owns the workflow. That's a falsifiable claim — it fails if LLM coding quality plateaus below the threshold where developers trust autonomous execution, or if the IDE category gets absorbed by browser-based dev environments. The dependency that has to hold is continued improvement in multi-file reasoning accuracy, and the trend line — model capability on SWE-bench style tasks improving roughly 2x per year — is still running. The second-order effect nobody is talking about: Background Agents create a new power asymmetry inside engineering teams, where the developer who knows how to write effective agent prompts becomes dramatically more productive than one who doesn't, which reshapes hiring and seniority definitions faster than most eng managers expect. Cursor is early to the 'agent as first-class editor citizen' framing and that's the right place to be on this curve.”

“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 clear: engineering teams at mid-market and enterprise companies where CISOs need audit trails before they'll approve AI tooling — that's a real procurement unlock and Cursor shipped exactly the right feature at the right time with audit logs. The pricing architecture scales with seat count, which aligns with value since more engineers means more agent usage, but the real expansion lever is whether teams move from individual Pro licenses to org-wide Business contracts, and the audit-log feature is the wedge for that exact motion. The moat question is harder: Cursor's defensibility is editor-layer context, but JetBrains and Microsoft both have that same layer and significantly more enterprise distribution. What would need to be true for this to win is that developer preference overrides IT procurement preference — which has happened before with tools like Slack, so it's not impossible. The business survives a 10x model price drop because their cost is inference and their value is workflow integration; that's the right structure.”