Skip to content
Nx IDEDocs

AI Providers & Models3 min read

Anthropic

Claude models with prompt caching

Anthropic’s Claude models are a first-class choice for Nexgile Code: strong agentic coding, long context windows, universal prompt-caching support across the roster, and fine-grained thinking controls. If you want one provider that handles everything from quick edits to day-long refactors, start here.

Get an API key

Section titled “Get an API key”
  1. Sign in at the Anthropic Console — console.anthropic.com.
  2. Open the API keys section and create a new key. (When no key is configured yet, the settings panel shows a Get Anthropic API Key button that takes you straight to the console’s keys page.)
  3. Add billing or credits to the Anthropic account; the key will not serve requests without it.

Configure in Nx IDE

Section titled “Configure in Nx IDE”
  1. Open Settings → Providers.
  2. Choose Anthropic in the API Provider dropdown.
  3. Paste your key into Anthropic API Key. Keys are stored in the editor’s Secret Storage, never in settings files.
  4. Pick a model and save.

Optional controls on the same panel:

  • Use custom base URL — route requests through a proxy or regional endpoint instead of https://api.anthropic.com. When enabled you can also toggle Pass Anthropic API Key as Authorization header instead of X-Api-Key, which some corporate gateways require.
  • Enable 1M context window (Beta) — shown for claude-sonnet-4-20250514, claude-sonnet-4-5, claude-sonnet-4-6, and claude-opus-4-6. Extends those models from 200K to 1M tokens of context at tiered pricing.

Claude models

Section titled “Claude models”

The default model is claude-sonnet-4-5. Current-generation options:

Model id Context Max output Thinking
claude-opus-4-8 1M (built in, flat pricing) 128K Adaptive thinking with effort levels
claude-opus-4-7 1M (built in, flat pricing) 128K Adaptive thinking with effort levels
claude-opus-4-6 200K (1M beta toggle) 128K Thinking budget
claude-sonnet-4-6 200K (1M beta toggle) 64K Thinking budget
claude-sonnet-4-5 200K (1M beta toggle) 64K Thinking budget
claude-haiku-4-5-20251001 200K 64K Thinking budget

Previous generations remain selectable for compatibility: claude-sonnet-4-20250514, claude-opus-4-5-20251101, claude-opus-4-1-20250805, claude-opus-4-20250514, claude-3-7-sonnet-20250219 (plus its always-thinking :thinking variant), claude-3-5-sonnet-20241022, claude-3-5-haiku-20241022, claude-3-opus-20240229, and claude-3-haiku-20240307.

Every model in the list supports prompt caching, and all of them accept image input except claude-3-5-haiku-20241022.

Prompt caching

Section titled “Prompt caching”

Claude prompt caching is applied automatically — no toggle needed. Nexgile Code’s system prompt, rules files, and conversation history are stable between turns, which makes agent traffic ideal cache material: cache reads are billed at a small fraction of the normal input price (for example $0.30 per million tokens instead of $3.00 on Sonnet-class models). Over a long task, most of your input tokens become cheap cache hits. The cost figure in the task header already accounts for cache writes and reads.

Thinking controls

Section titled “Thinking controls”

Claude models expose two different reasoning styles, and the settings panel adapts to whichever the selected model supports:

  • Adaptive thinking (claude-opus-4-8, claude-opus-4-7): the model decides how long to think based on the request; you steer it with Model Reasoning Effort — None, Low, Medium, High, Extra High, or Max (the default for these models). There is no manual token budget; these models reject one.
  • Thinking budgets (Sonnet 4.x, Opus 4.6 and earlier 4-series, Haiku 4.5, and the 3.7 :thinking variant): enable reasoning and set Max Thinking Tokens to cap how much the model deliberates before answering.

See Model Behavior & Costs for how these controls interact with output limits and pricing.

Section titled “Recommended uses”
  • Opus 4.8 / 4.7 — architecture work, multi-hour autonomous tasks, gnarly debugging, and anything where you want the 1M-token context without beta toggles. Pin one to Architect or Orchestrator mode.
  • Sonnet 4.5 / 4.6 — the everyday driver: excellent code quality at a third of Opus input pricing. A natural default for Code and Debug modes.
  • Haiku 4.5 — high-volume, low-stakes work: commit messages, quick explanations, bulk mechanical edits, or a dedicated condensing profile.
Suggested profile pairing:


  opus-deep     claude-opus-4-8    pinned to Architect
  sonnet-daily  claude-sonnet-4-5  pinned to Code / Debug
  haiku-cheap   claude-haiku-4-5-20251001   condensing profile
Section titled “Related”