skawld-sdk  by skawld

TypeScript agent harness for Node.js and Bun

Created 2 months ago
303 stars

Top 87.9% on SourcePulse

GitHubView on GitHub
Project Summary

Skawld Agent SDK

Skawld Agent SDK offers a powerful, open-source TypeScript harness for embedding sophisticated AI agent functionalities directly into Node.js and Bun applications. It streamlines the integration of a complete agent loop—encompassing tools, session management, granular permissions, real-time event streaming, and subagent orchestration—via a single import. This SDK targets developers aiming to rapidly infuse their applications with advanced AI capabilities, providing a flexible and robust framework for building intelligent agents.

How It Works

The SDK employs a modular design, enabling developers to instantiate an Agent configured with a specific LLM provider (e.g., Anthropic, OpenAI) and a curated set of tools. It manages the core agent execution flow, including message parsing, tool invocation, and event propagation. Notable features include a configurable permission engine, persistent session storage (defaulting to SQLite, with an in-memory option), and support for hierarchical subagents, facilitating complex multi-agent interactions. This approach prioritizes developer experience and extensibility.

Quick Start & Requirements

  • Installation: Utilize your preferred package manager: bun add @skawld/agent-sdk, npm install @skawld/agent-sdk, pnpm add @skawld/agent-sdk, or yarn add @skawld/agent-sdk.
  • Prerequisites: Requires Node.js 18+ or Bun 1.1+. The SDK is ESM-only.
  • Environment Variables: API keys are necessary for providers: ANTHROPIC_API_KEY (AnthropicProvider) and OPENAI_API_KEY (OpenAI providers). Optional configuration overrides include SKAWLD_MODEL and SKAWLD_CONFIG_DIR.
  • Documentation: Comprehensive documentation is available at https://skawld.com/docs.
  • Examples: Runnable examples, such as minimal-agent.ts and interactive-cli.ts, are provided within the repository.

Highlighted Details

  • Integrates with multiple LLM providers, including Anthropic and OpenAI (Chat Completions, Responses APIs).
  • Features a robust event streaming system for real-time feedback on agent operations and subagent activity.
  • Provides built-in tools and helpers for common agent tasks, with clear pathways for custom tool integration.
  • Offers flexible session persistence options, defaulting to SQLite for state management across application restarts.

Maintenance & Community

No specific details regarding notable contributors, sponsorships, or community channels (e.g., Discord, Slack) were present in the provided README content.

Licensing & Compatibility

  • License: MIT.
  • Compatibility: The permissive MIT license generally permits commercial use and integration within closed-source applications without significant restrictions.

Limitations & Caveats

  • The SDK strictly enforces an ESM module format, necessitating compatible build configurations or runtimes.
  • Successful integration with specific LLM providers depends on correct API key setup and environment variable configuration.
Health Check
Last Commit

1 month ago

Responsiveness

Inactive

Pull Requests (30d)
2
Issues (30d)
0
Star History
7 stars in the last 30 days

Explore Similar Projects

Starred by Chip Huyen Chip Huyen(Author of "AI Engineering", "Designing Machine Learning Systems"), Elvis Saravia Elvis Saravia(Founder of DAIR.AI), and
1 more.

agentkit by BCG-X-Official

0%
2k
Starter kit for building constrained agents
Created 2 years ago
Updated 4 months ago
Feedback? Help us improve.