@effect/ai-anthropic

Anthropic provider integration for Effect AI SDK

An Anthropic provider integration that provides type-safe, Effect-based access to Anthropic’s Claude API, including messages, streaming, and tool calling.

Installation

pnpm add @effect/ai-anthropic effect

Setup

Create an Anthropic client by providing your API key:

import { AnthropicClient } from "@effect/ai-anthropic"
import { Effect, Layer, Redacted } from "effect"

const AnthropicLive = AnthropicClient.layer({
  apiKey: Redacted.make(process.env.ANTHROPIC_API_KEY!)
})

Language Model

Use the Anthropic language model for text generation:

import { AnthropicLanguageModel } from "@effect/ai-anthropic"
import { Effect } from "effect"
import { LanguageModel } from "effect/unstable/ai"

const program = Effect.gen(function*() {
  const result = yield* LanguageModel.generate({
    prompt: "Explain the concept of monads in functional programming"
  })
  console.log(result)
}).pipe(
  Effect.provide(AnthropicLanguageModel.model("claude-sonnet-4-20250514")),
  Effect.provide(AnthropicLive)
)

Effect.runPromise(program)

Streaming Responses

Stream responses for real-time text generation:

import { AnthropicLanguageModel } from "@effect/ai-anthropic"
import { Effect, Stream } from "effect"
import { LanguageModel } from "effect/unstable/ai"

const program = Effect.gen(function*() {
  const stream = yield* LanguageModel.streamText({
    prompt: "Write a short story about a robot"
  })
  
  yield* Stream.runForEach(stream, (chunk) =>
    Effect.sync(() => process.stdout.write(chunk.text))
  )
}).pipe(
  Effect.provide(AnthropicLanguageModel.model("claude-sonnet-4-20250514")),
  Effect.provide(AnthropicLive)
)

Effect.runPromise(program)

Tool Calling

Use Claude’s native tools or define custom tools:

import { AnthropicLanguageModel, AnthropicTool } from "@effect/ai-anthropic"
import { Effect, Schema } from "effect"
import { LanguageModel, Tool, Toolkit } from "effect/unstable/ai"

// Use Anthropic's native Bash tool
const program = Effect.gen(function*() {
  const toolkit = Toolkit.make(AnthropicTool.Bash)
  
  const result = yield* LanguageModel.generate({
    prompt: "List all TypeScript files in the current directory",
    toolkit
  })
  
  console.log(result)
}).pipe(
  Effect.provide(AnthropicLanguageModel.model("claude-sonnet-4-20250514")),
  Effect.provide(AnthropicLive)
)

// Define a custom tool
const SearchTool = Tool.make("SearchTool", {
  description: "Search for files matching a pattern",
  parameters: Schema.Struct({ pattern: Schema.String }),
  success: Schema.String
})

const customProgram = Effect.gen(function*() {
  const toolkit = Toolkit.make(SearchTool)
  const toolkitLayer = toolkit.toLayer({
    SearchTool: ({ pattern }) => Effect.succeed(`Found: ${pattern}`)
  })
  
  const result = yield* LanguageModel.generate({
    prompt: "Search for all markdown files",
    toolkit
  })
  
  return result
}).pipe(
  Effect.provide(AnthropicLanguageModel.model("claude-sonnet-4-20250514")),
  Effect.provide(toolkitLayer),
  Effect.provide(AnthropicLive)
)

Native Tools

Anthropic provides several native tools:

• AnthropicTool.Bash: Execute bash commands
• AnthropicTool.CodeExecution: Run code in a sandbox
• AnthropicTool.ComputerUse: Interact with a computer interface
• AnthropicTool.Memory: Store and retrieve information across messages
• AnthropicTool.TextEditor: Edit text files

Error Handling

Handle Anthropic-specific errors:

import { AnthropicError } from "@effect/ai-anthropic"
import { Effect } from "effect"

const program = Effect.gen(function*() {
  // Your Anthropic API calls
}).pipe(
  Effect.catchTag("HttpClientError", (error) =>
    Effect.sync(() => {
      console.error("Anthropic API error:", error)
      // Handle specific error cases
    })
  )
)

Telemetry

Integrate with OpenTelemetry for observability:

import { AnthropicTelemetry } from "@effect/ai-anthropic"
import { Effect } from "effect"

// Telemetry attributes are automatically added to traces
// when using the Anthropic client with OpenTelemetry integration

API Modules

• AnthropicClient: HTTP client for Anthropic API
• AnthropicConfig: Configuration options
• AnthropicError: Error type augmentation
• AnthropicLanguageModel: Language model implementation
• AnthropicTelemetry: OpenTelemetry integration
• AnthropicTool: Provider-defined tools

Related Packages

• @effect/ai-openai
• @effect/ai-openrouter
• @effect/opentelemetry