@elizaos/plugin-anthropic

A multi-language Anthropic Claude API client for elizaOS, providing text generation and structured JSON object generation capabilities.

🌐 Multi-Language Support

This plugin is implemented in three languages for maximum flexibility:

| Language | Package | Registry |
| ---------- | --------------------------- | --------- |
| TypeScript | @elizaos/plugin-anthropic | npm |
| Rust | elizaos-plugin-anthropic | crates.io |
| Python | elizaos-plugin-anthropic | PyPI |

All implementations share the same API design and behavior.

Features

- 🚀 Text Generation - Generate text with Claude models (small/large)
- 📋 Object Generation - Generate structured JSON objects with validation
- 🔒 Strong Types - No any or unknown types, full type safety
- ⚡ Fail Fast - Immediate errors on invalid input, no silent failures
- 🧪 Real Integration Tests - Tests against live Anthropic API

Quick Start

$3

``typescript
import { anthropicPlugin } from "@elizaos/plugin-anthropic";
import { AgentRuntime, ModelType } from "@elizaos/core";

// Register the plugin
const runtime = new AgentRuntime({
plugins: [anthropicPlugin],
});

// Generate text
const text = await runtime.useModel(ModelType.TEXT_LARGE, {
prompt: "Explain quantum computing in simple terms",
});

// Generate JSON object
const result = await runtime.useModel(ModelType.OBJECT_SMALL, {
prompt: "Create a user profile with name, email, and age",
schema: { type: "object" },
});
`

$3

`rust
use elizaos_plugin_anthropic::{AnthropicClient, AnthropicConfig, TextGenerationParams};

#[tokio::main]
async fn main() -> anyhow::Result<()> {
let config = AnthropicConfig::from_env()?;
let client = AnthropicClient::new(config)?;

let params = TextGenerationParams::new("Explain quantum computing")
.with_max_tokens(1024)
.with_temperature(0.7);

let response = client.generate_text_large(params).await?;
println!("Response: {}", response.text);
Ok(())
}
`

$3

`python
import asyncio
from elizaos_plugin_anthropic import AnthropicClient, AnthropicConfig

async def main():
config = AnthropicConfig.from_env()
async with AnthropicClient(config) as client:
response = await client.generate_text_large("Explain quantum computing")
print(f"Response: {response.text}")

asyncio.run(main())
`

Installation

$3

`bash
npm install @elizaos/plugin-anthropic


or


bun add @elizaos/plugin-anthropic
`

$3

`toml
[dependencies]
elizaos-plugin-anthropic = "1.0"
`

$3

`bash
pip install elizaos-plugin-anthropic
`

Configuration

All implementations use the same environment variables:

| Variable | Required | Default | Description |
| ---------------------------------- | -------- | --------------------------- | --------------------------------------- |
| ANTHROPIC_API_KEY | Yes | - | Your Anthropic API key |
| ANTHROPIC_BASE_URL | No | https://api.anthropic.com | API base URL |
| ANTHROPIC_SMALL_MODEL | No | claude-3-5-haiku-20241022 | Small model ID |
| ANTHROPIC_LARGE_MODEL | No | claude-sonnet-4-20250514 | Large model ID |
| ANTHROPIC_TIMEOUT_SECONDS | No | 60 | Request timeout |
| ANTHROPIC_EXPERIMENTAL_TELEMETRY | No | false | Enable telemetry (TS only) |
| ANTHROPIC_COT_BUDGET | No | 0 | Chain-of-thought token budget (TS only) |

Available Models

| Model ID | Size | Description |
| ---------------------------- | ----- | -------------------- |
| claude-3-5-haiku-20241022 | Small | Fast and efficient |
| claude-sonnet-4-20250514 | Large | Most capable |
| claude-3-5-sonnet-20241022 | Large | Balanced performance |
| claude-3-opus-20240229 | Large | Previous flagship |

API Reference

$3

- TEXT_SMALL - Text generation with small model
- TEXT_LARGE - Text generation with large model
- OBJECT_SMALL - JSON generation with small model
- OBJECT_LARGE - JSON generation with large model

$3

| Parameter | Type | Description |
| --------------- | --------- | --------------------------------------------- |
| prompt | string | The prompt to generate from |
| system | string? | Optional system prompt |
| maxTokens | number? | Maximum tokens to generate |
| temperature | number? | Randomness (0-1, can't use with topP) |
| topP | number? | Nucleus sampling (can't use with temperature) |
| stopSequences | string[]? | Stop generation at these sequences |

$3

| Parameter | Type | Description |
| ------------- | ------- | ----------------------------------------------- |
| prompt | string | Description of the object to generate |
| schema | object? | Optional JSON schema |
| temperature | number? | Randomness (default: 0.2 for structured output) |

Project Structure

`
plugin-anthropic/
├── typescript/ # TypeScript implementation
│ ├── index.ts # Main entry point
│ ├── models/ # Model handlers
│ ├── providers/ # Anthropic client factories
│ ├── types/ # Type definitions
│ ├── utils/ # Utilities (config, JSON parsing)
│ └── __tests__/ # Unit and integration tests
├── rust/ # Rust implementation
│ ├── src/ # Source code
│ │ ├── lib.rs # Library entry
│ │ ├── client.rs # API client
│ │ ├── config.rs # Configuration
│ │ ├── models.rs # Model definitions
│ │ ├── types.rs # Type definitions
│ │ └── error.rs # Error types
│ └── tests/ # Integration tests
├── python/ # Python implementation
│ ├── elizaos_plugin_anthropic/
│ │ ├── __init__.py # Package entry
│ │ ├── client.py # API client
│ │ ├── config.py # Configuration
│ │ ├── models.py # Model definitions
│ │ ├── types.py # Type definitions
│ │ └── errors.py # Error types
│ └── tests/ # Integration tests
├── package.json # npm package config
└── README.md # This file
`

Development

$3

- TypeScript: Bun or Node.js 18+
- Rust: Rust 1.70+ with cargo
- Python: Python 3.11+

$3

`bash


TypeScript

cd typescript
npx vitest

With integration tests (requires API key)

ANTHROPIC_API_KEY=your-key npx vitest

Rust

cd rust
cargo test

With integration tests

ANTHROPIC_API_KEY=your-key cargo test -- --ignored

Python

cd python
pip install -e ".[dev]"
pytest

With integration tests

ANTHROPIC_API_KEY=your-key pytest -m integration
`

$3

`bash


TypeScript

bun run build

Rust (native library)

cd rust && cargo build --release

Rust (WASM)

cd rust && wasm-pack build --target web --out-dir pkg/web

Python (wheel)

cd python && pip install build && python -m build
`

Error Handling

All implementations follow a fail-fast philosophy:

- No try-catch blocks that swallow errors
- No fallback modes or mock modes
- Immediate validation of all inputs
- Clear error messages with actionable information

$3

| Error | Description |
| ----------------------- | ---------------------------------- |
| ApiKeyError | API key missing or invalid |
| ConfigError | Configuration problem |
| RateLimitError | Rate limit exceeded (retryable) |
| ApiError | API returned an error |
| JsonGenerationError | Failed to parse JSON from response |
| InvalidParameterError | Invalid parameter provided |
| NetworkError | Network connectivity issue |
| TimeoutError` | Request timed out |

License

MIT - see LICENSE

Contributing

See the elizaOS contributing guide.