@llm-ports/adapter-anthropic

Direct @anthropic-ai/sdk adapter for llm-ports. Implements LLMPort for Claude models with prompt caching, vision, and tool use.

Install

pnpm add @llm-ports/core @llm-ports/adapter-anthropic @anthropic-ai/sdk zod

Configure

import { createRegistryFromEnv } from "@llm-ports/core";
import { createAnthropicAdapter } from "@llm-ports/adapter-anthropic";

const registry = createRegistryFromEnv({
  adapters: {
    anthropic: createAnthropicAdapter({
      apiKey: process.env.ANTHROPIC_API_KEY!,
    }),
  },
});

const llm = registry.getPort();

const result = await llm.generateText({
  taskType: "triage",
  prompt: "Classify this email: ...",
});

.env:

LLM_PROVIDER_FAST=anthropic|claude-haiku-4-5|cost:5/day
LLM_PROVIDER_SMART=anthropic|claude-sonnet-4-6-20250514|cost:50/day
LLM_TASK_ROUTE_TRIAGE=fast,smart

Adapter options

interface AnthropicAdapterOptions {
  apiKey: string;
  baseURL?: string;                            // typically only useful for testing
  fetch?: typeof fetch;                        // inject custom fetch (tests, proxies)
  validationStrategy?: ValidationStrategy;
  pricingOverrides?: Record<string, ModelPricing>;
  imageSizeLimitBytes?: number;                // default 5 MB (Anthropic's per-image limit)
  dangerouslyAllowBrowser?: boolean;           // opt in to browser execution (alpha.9)
  onRetry?: OnRetry;                           // observability hook for retries
}

dangerouslyAllowBrowser (alpha.9+)

The Anthropic SDK refuses to construct in a browser environment unless dangerouslyAllowBrowser: true is passed explicitly. When enabled, the SDK auto-adds the anthropic-dangerous-direct-browser-access header on every request. Set this option only when the API key is NOT a long-lived secret: short-lived proxy tokens, BYO-key UIs where the end user supplies their own key, or trusted internal tools running behind auth. For server-side proxy patterns where the secret stays on the server, leave it unset.

const adapter = createAnthropicAdapter({
  apiKey: ephemeralUserKey,
  dangerouslyAllowBrowser: true,
});

Bundled pricing

The bundled ANTHROPIC_PRICING table covers Claude Opus 4.5, Sonnet 4.5/4.6, Haiku 4.5 with full cache-read / cache-write rates. Override per model via pricingOverrides.

Supported features

| Feature | Status | |---------|--------| | generateText | ✓ | | generateStructured (Zod schemas) | ✓ with retry-with-feedback (default) | | streamText | ✓ (yields text chunks) | | streamStructured (partial JSON) | ✓ (best-effort partial parse) | | runAgent (multi-turn tool use) | ✓ | | Prompt caching | ✓ — reported in cost via cacheReadTokens / cacheWriteTokens | | Vision input — base64 images | ✓ | | Vision input — URL images | ✓ | | Audio input | ✗ — throws ContentBlockUnsupportedError | | Embeddings | ✗ — Anthropic ships no embedding models | | AbortSignal cancellation | ✓ entry + in-flight (alpha.6) |

Content blocks supported

text, image (base64 + URL), tool_use, tool_result. Throws ContentBlockUnsupportedError for audio.

Cancellation

Full AbortSignal support shipped in 0.1.0-alpha.6. The signal is threaded into both client.messages.create (non-streaming) and client.messages.stream, so controller.abort() cancels the in-flight HTTP request. runAgent also re-checks the signal between steps so cancellation propagates mid-loop. See the Cancellation guide.

Temperature handling on reasoning models

Claude 4.5+ reasoning models (claude-opus-4-5*, claude-sonnet-4-5*) reject custom temperature values. The adapter learns this constraint at runtime (or via the static KNOWN_TEMPERATURE_REJECTORS catalog) and strips the parameter automatically. The onRetry hook fires once with reason: "capability-fallback" + capability: "temperatureLocked". See docs/known-quirks.md for the full catalog.

SDK version compatibility

The adapter is tested against @anthropic-ai/sdk 0.32.0 ≤ v < 0.50.0. If your installed SDK is outside that range, a one-time console.warn fires at adapter creation with guidance.

Reading next

• Anthropic adapter docs — full feature deep-dive
• Known model quirks — temperature rejection, prompt-caching nuances
• Tool-use security guide — runAgent safety patterns
• Multi-provider routing — chain Anthropic with OpenAI / Gemini fallbacks