@arnilo/prism-provider-anthropic

Native Anthropic Messages (/v1/messages) provider for Prism. No vendor SDK.

import { createAnthropicProviderPackage, listAnthropicModels } from "@arnilo/prism-provider-anthropic";

api.registerProviderPackage(createAnthropicProviderPackage({ apiKey: "fake-anthropic-key" }));

// Caller-gated discovery (never during setup)
const models = await listAnthropicModels({ apiKey: "fake-anthropic-key" });
api.registerProviderPackage(createAnthropicProviderPackage({ apiKey: "fake-anthropic-key", models }));

Exports:

• createAnthropicProviderPackage()
• createAnthropicMessagesProvider()
• listAnthropicModels() / mapAnthropicModel() / defineAnthropicModel()
• anthropicModels (featured offline aliases)
• anthropicThinking / anthropicEffort / anthropicPreserveThinking
• applyAnthropicCacheControl / anthropicCacheEnabled

Security defaults:

• No network calls during import, setup, build, or default tests.
• No automatic environment, file, keychain, or shell credential lookup.
• Credentials are resolved per request from caller-supplied values or resolvers; this package registers api_key only.
• No Claude Code/Claude.ai subscription OAuth, setup-token, or credential-file import. Anthropic requires third-party products to use API keys or supported cloud providers: https://docs.anthropic.com/en/docs/claude-code/legal-and-compliance.
• Provider-owned headers (content-type, x-api-key, anthropic-version) win over caller headers.
• Live tests stay opt-in behind PRISM_LIVE_PROVIDER_TESTS=1 + ANTHROPIC_API_KEY.

Featured models (offline bootstrap): claude-opus-4-8, claude-sonnet-5, claude-haiku-4-5, claude-fable-5.

Cache: featured models use cache.kind: "cache_control"; markers apply only to selected cache.breakpoints ("long" → ttl: "1h").

Docs: docs/providers/anthropic.md