> For the complete documentation index, see [llms.txt](https://kai-7.gitbook.io/axiomkit/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://kai-7.gitbook.io/axiomkit/about-axiomkit.md).

# About AxiomKit

TypeScript framework for building autonomous AI agents with unified LLM provider support.

## What is AxiomKit?

* Brief, high-level overview of AxiomKit as a next-generation framework for building autonomous AI agents.
* Mention its core principles: Modularity, Extensibility, Scalability, and advanced Agent Autonomy.

## Why AxiomKit?

In a rapidly evolving landscape of AI agent frameworks, AxiomKit distinguishes itself by pushing the boundaries of what autonomous agents can achieve, coupled with an unparalleled developer experience.

* **TypeScript Compatible** - Full Type Safety and developer easier to use.
* **Model-agnostic -** worked with multiple latest models like Open AI , DeepSeek , and any others AI SDK Providers.
* **Agent Autonomy & Advanced Reasoning:** 
  * Dynamic Re-planning: Unlike systems that merely re-attempt failed actions, AxiomKit agents are equipped with advanced planning algorithms. If a plan falters or the environment shifts, our agents can intelligently re-evaluate the situation, formulate entirely new strategies, and even reformulate their primary goals to adapt to emergent circumstances, minimizing human oversight.
  * Built-in Self-Correction & Self-Improvement Loops: AxiomKit agents aren't just intelligent; they're *learning*. The framework provides native mechanisms for agents to reflect on their past actions, identify inefficiencies, and automatically refine their internal logic, decision-making processes, or prompt strategies.
* Memory Flexible: Our memory system is not just a database; it's a living archive. Agents can perform highly precise temporal and contextual queries, such as "What did I discuss with User X last Tuesday regarding Project Y, but exclude messages from Channel Z?" This allows for highly relevant and accurate information retrieval, critical for long-running, personalized agent interactions.
* Scale Plugin: Scale by integrate external plugin production for twitter , discord ,telegram and more.

AxiomKit is built for developers, researchers, and organizations who demand unparalleled intelligence, robust capabilities, and a streamlined development workflow for their next-generation AI agent initiatives.

## **How AxiomKit Work?**

* Core Runtime: The central nervous system of AxiomKit, responsible for managing the agent's lifecycle, handling event loops, and orchestrating interactions between different components.
* Agents: The intelligent entities built within the framework, each with its own goals, context, and capabilities. An agent leverages the other components to perceive, reason, and act.
* LLM Providers: Abstractions that allow agents to seamlessly integrate with various Large Language Models (LLMs) – whether cloud-based APIs (OpenAI, Anthropic, Google Gemini, Groq) or local models (Ollama, Llama.cpp).
* Memory System: A pluggable and advanced memory layer that provides both short-term (working context) and long-term (persistent knowledge base, including multi-modal data) recall.
* Actions (Tools): The defined capabilities an agent can execute, ranging from simple function calls (e.g., sending an email) to complex external API interactions (e.g., querying a database, interacting with a blockchain).
* Plugins: Self-contained packages that bundle related actions, memory adapters, and other functionalities, allowing for easy extension of the framework's capabilities (e.g., a "Gmail Plugin" for email management, a "Twitter Plugin" for social media interaction).
* Input & Output Adapters: Modules that enable agents to receive information from and send information to various external platforms (e.g., Discord, Slack, Webhooks, custom APIs).
* Contexts: Isolated environments or sessions that allow agents to maintain continuity and manage separate interactions or long-running tasks without interference.