Agentica
TypeScript AI AI Function Calling Framework enhanced by compiler skills.
Last verified:
What is Agentica?
Agentica is an open-source Agentic AI framework specialized in LLM function calling, developed by Wrtn Technologies. It transforms TypeScript classes and Swagger/OpenAPI specifications into fully functional AI agents instantly, without requiring any machine learning expertise. The framework leverages the TypeScript compiler to automatically parse classes, read JSDoc comments, and generate LLM function call schemas that AI models can understand and use accurately.
Key features include compiler-driven development that automatically constructs function calling schemas without manual coding, automatic JSON schema conversion that handles specification differences between LLM vendors (OpenAI, Google, Anthropic, DeepSeek, Meta), validation feedback that detects and corrects AI mistakes in argument composition, and a Selector Agent that filters candidate functions to minimize context usage and reduce token consumption. Agentica supports three protocols: TypeScript Class, Swagger/OpenAPI Document, and MCP (Model Context Protocol) Server.
Agentica is designed for TypeScript developers, backend developers, and anyone who can write functions. It includes over 300 built-in connectors (GitHub, Slack, Reddit, Google Calendar, AWS S3, Calendly, DALL-E-3, and more), vector store support for RAG, and a CLI setup wizard. The framework scales from hobby projects to enterprise-level applications, as demonstrated by the @samchon/shopping-backend with 289 API functions for e-commerce. It supports NodeJS, NestJS, React, and standalone application types with WebSocket protocol for client-server communication.
Agentica pricing
Pricing model: Freemium
Agentica is completely free and open-source under the MIT License. Users can use, modify, distribute, and sell the software without restriction. The framework itself has no licensing fees or paid tiers. However, users must provide their own LLM API keys (such as OpenAI API key) which are billed separately by the respective AI model providers. The setup wizard requires entering an OPENAI_API_KEY during project creation. No subscription plans or paid features exist for Agentica itself.
Agentica pros
- No ML knowledge required to build AI agents
- Turns TypeScript classes into AI agents instantly
- Converts OpenAPI/Swagger specs to agents automatically
- Compiler-driven development eliminates manual schema writing
- Automatic JSON schema conversion for all major LLM vendors
- Validation feedback corrects AI mistakes automatically
- Over 300 built-in connectors available
- Open-source under permissive MIT License
- Simple CLI setup wizard with project templates
- Scales from hobby to enterprise projects
- Supports multi-step call coordination automatically
- Reduces token consumption with Selector Agent
- 300+ embedded controllers (GitHub, Slack, Calendar, etc.)
- Vector store support for RAG applications
- Works with ChatGPT, Gemini, Claude, DeepSeek, and Llama
- No complex agent graph or workflow needed
- Document-driven development paradigm
Agentica cons
- Requires OpenAI API key (paid service)
- Standalone frontend not recommended for production
- yarn berry package manager not supported
- Requires TypeScript knowledge to use effectively
- Backend API authentication must be handled separately
- Learning curve for Document Driven Development paradigm
- Requires backend server for production deployments
- Verbose setup code with multiple controller configurations
- Primarily optimized for TypeScript/Node.js ecosystem
- Benchmarking module requires separate setup for quality assurance
Frequently asked questions about Agentica
What is Agentica?
Agentica is an open-source Agentic AI framework specialized in LLM function calling. It turns your TypeScript classes and Swagger/OpenAPI specifications into real AI agents instantly without requiring machine learning expertise. Simply list the functions you need, and Agentica creates the agent automatically.
Do I need ML knowledge to use Agentica?
No, you don't need any ML or AI expertise. If you can write TypeScript functions or have a well-documented OpenAPI spec, you can build AI agents. Agentica leverages the TypeScript compiler to handle all the tedious schema generation work automatically.
What protocols does Agentica support?
Agentica supports three protocols: TypeScript Class (convert TypeScript classes directly), Swagger/OpenAPI Document (use existing API documentation), and MCP (Model Context Protocol) Server. You just prepare functions from any of these protocols and Agentica creates the agent.
Which AI models work with Agentica?
Agentica works with ChatGPT (OpenAI), Gemini (Google), Claude (Anthropic), DeepSeek, and Llama (Meta). It automatically handles specification differences between LLM vendors through JSON schema conversion, ensuring seamless integration regardless of your chosen AI model.
Is Agentica free to use?
Yes, Agentica is completely open-source and available under the MIT License. You can use it freely for personal, commercial, or enterprise projects without any licensing fees. However, you need to provide your own LLM API keys which are billed separately by the AI model providers.
How do I set up an Agentica project?
Run the command 'npx agentica start <directory>' to launch the Agentica Setup Wizard. The CLI will ask you to choose a package manager (npm, pnpm, yarn, bun), project type (NodeJS, NestJS, React, or Standalone), and embedded controllers (GitHub, Slack, Reddit, etc.). Finally, enter your OpenAI API key to complete setup.
What are embedded controllers?
Embedded controllers are pre-built functions you can use in your agent, such as Google Calendar, GitHub, Reddit, Slack, AWS S3, Calendly, DALL-E-3, and Arxiv Search. During project setup, you can select multiple controllers to embed in your agent without writing code from scratch.
Can Agentica handle enterprise-level projects?
Yes, Agentica is suitable for enterprise-level projects. The @samchon/shopping-backend demonstrates this with 289 API functions covering e-commerce features like product search, shopping cart, orders, payment, and refunds. It follows Document Driven Development paradigm and includes benchmarking for quality assurance.
How does Agentica handle AI mistakes?
Agentica provides validation feedback that detects and corrects AI mistakes in argument composition. It automatically generates feedback when the AI makes errors, helping it learn and recover automatically. This dramatically reduces errors and improves reliability compared to vanilla function calling.
What is the difference between Agentica and agent workflow frameworks?
Agentica uses function-driven development which is general-purpose, easy, stable, and flexible. Agent workflow frameworks are specific-purpose, difficult, but stable yet inflexible. Agentica resolves the difficulty and instability problems of traditional function calling, making it easier than workflow-based approaches while maintaining flexibility for general use cases.