Typescript Agent Toolkit
Null Shot's Typescript Vibework (Framework) for Serverless AI Agents
Ask AI about Typescript Agent Toolkit
Powered by Claude Β· Grounded in docs
I know everything about Typescript Agent Toolkit. Ask me about installation, configuration, usage, or troubleshooting.
0/500
Reviews
Documentation
π NullShot - Typescript Agent Framework
Building the future of AI Agent Interoperability
Pre-Alpha: This project is in active development.
Welcome to a new paradigm in AI development. MCP provides the foundation for building intelligent, interoperable agents that can communicate, evolve, and generate value at the edge of the network.
Vision
We're extending Cloudflare's vision for AI Agents with a focus on web3 and MCPs as plugins:
- π€ AI Agents as teammates/organizations generating revenue and performing advanced operations
- π° Cost-effective shared hosting options
- π Secure sensitive assets (trading agents, treasuries, etc.)
- π Self-improving agents based on collective usage
- πΈ Drive MCP usage revenue to open source contributors
- πΌ Monetization avenues for closed-source MCP use cases
- βοΈ Seamless configuration options
- π Quick iteration on ideas locally and in-cloud
- π No vendor lock-in, self-hosting and personal account options
Project Status
This project is in pre-alpha and actively evolving. Current focus areas:
Ready for Use
- β Core MCP Framework
- β Multi Session & Authentication Patterns
- β Official MCP WebSocket Support and HTTP Streaming Support
- β Agent Framework (AI SDK)
- β Seamless MCP Plugins (mcp.json) for Agents
- β Agent MCP Dependency Management
- β MCP Webhook / External Service Support
- β Playground UI for LLMs + MCPs
In Development
- βοΈ Cloudflare Service Examples (KV, D1, DO state, Analytics, Workflows, Schema Migrations)
- π€ LangChain and Agent SDK examples coming soon
- π Cloudflare Pages (SSE / Fullstack) Examples
- π Authentication (OAuth, JWT)
Quick Start
Get started with the Null Shot CLI to create MCP servers and AI agents:
Install the CLI
npm install -g @nullshot/cli
Create a new MCP server
nullshot create mcp
Create a new Agent
nullshot create agent
Install MCP dependencies
nullshot install
Run in development mode
nullshot dev
Sign in and inspect Jam conversations
nullshot login
nullshot jam
nullshot messages <room-id> --full
nullshot logs <room-id>
nullshot errors <room-id>
π€ AI Provider Support
This framework supports 6 major AI providers with dynamic model fetching and official SDK integration:
Supported Providers
| Provider | Latest Models | SDK | Dynamic | API Key |
|---|---|---|---|---|
| OpenAI | GPT-4o, GPT-4o-mini, GPT-3.5-turbo | @ai-sdk/openai | β | β |
| Anthropic | Claude Opus 4.1, Claude Sonnet 4, Claude 3.7 | @ai-sdk/anthropic | β | β |
| DeepSeek | DeepSeek-Chat, DeepSeek-Coder | @ai-sdk/deepseek | β | β |
| Workers AI | Llama 3.1/3.2, Gemma 2, Mistral 7B (~49 models) | workers-ai-provider | β | Cloudflare |
| Gemini | Gemini 1.5 Pro Latest, Gemini 1.5 Flash Latest | @ai-sdk/google | β | β |
| Grok | Grok-4, Grok-3, Grok-3-mini, Grok-2-1212 | @ai-sdk/xai | β | β |
Key Features
- π Dynamic Model Fetching: Real-time model lists from provider APIs with intelligent fallbacks
- π― Official SDK Integration: Uses official AI SDK providers for consistent, reliable integration
- πΎ Smart Caching: Model lists cached for 30 minutes with API key validation
- π§ Provider-Aware UI: Auto-loading saved API keys and intelligent model selection
- π‘ Streaming Support: Real-time responses across all providers
- β‘ Fallback Handling: Graceful degradation when APIs are unavailable
Implementation Examples
- Next.js Web App:
examples/playground-showcase- Full-featured chat interface - Cloudflare Workers Agent:
examples/simple-prompt-agent- Server-side AI agent - React Components:
packages/playground- Reusable UI components
Documentation
Comprehensive documentation is available at Null Shot Docs:
- Project Overview - Get started with Null Shot
- Agent Framework - Getting Started - Build AI agents with Cloudflare Workers
- MCP Framework Overview - Model Context Protocol implementation
- Platform Overview - Understanding the platform architecture
- Common Services - Cloudflare services integration
- Playground - Interactive development environment
Release Process
This repository uses an automated release workflow following semantic versioning:
- Pull Request Testing - When you create a PR, it automatically runs tests and a semantic-release dry run
- Automated Publishing - When merged to main, changed packages are automatically published to npm
- Versioning - Package versions are determined by Conventional Commits standards
For detailed information about our release process, see .github/RELEASE_PROCESS.md.
Contributing
We welcome contributions! Our vision is to create a collaborative ecosystem where AI and human developers work together. Soon, we'll have an AI agent to audit and govern contributions based on our shared vision.
If you're interested in contributing, please:
- Join our Discord community
- Watch this repository for updates
- Star the project if you find it interesting
License
MIT License - see the LICENSE file for details.
