Agentic AI Kit is an open-source Python framework for building AI agents with tools, structured reasoning, retrieval, adaptive memory and data intelligence. The framework combines practical agent engineering with experimental work on emotion-aware episodic memory: systems that do more than retrieve past events and instead model how experience can influence retention, identity, expectations and future action.
Install Agentic AI Kit
View on GitHub
One framework. Multiple agent capabilities.
Create conversational agents, tool-using agents, structured-output agents, analytical agents and retrieval-augmented systems using a consistent Python foundation.
Use short-term, long-term and shared memory components, or explore an emotion-aware episodic memory architecture designed around appraisal, retention, identity and policy.
Use SQL-like AI functions to generate, summarize, classify and extract information from rows while preserving a familiar relational workflow.
Many agent frameworks begin with orchestration. Agentic AI Kit begins with a more fundamental question:
What should an intelligent system remember, and how should that memory change what it does next?
The project treats memory as an operational component rather than a transcript archive. Experiences can be appraised, retained, corrected, retrieved and used as evidence for future decisions. Around that research core, the kit provides practical components for tools, RAG, structured output, multi-agent coordination, machine-learning workflows, MCP-style integrations and local SQL AI enrichment.
Core capabilities
Model identity alignment, outcome evidence, expectation violation, salience, recurrence and retention. Maintain an auditable journal so corrected or deleted evidence does not continue influencing future decisions.
Add generative functions to SQL-style workflows. Generate text, summarize records, classify rows or extract retrieval signals without abandoning the structure of a query.
Build simple conversational agents or specialized agents for tools, structured output, reasoning, analysis, machine learning and retrieval-augmented generation.
Register Python functions as tools and allow agents to select and invoke capabilities when needed.
Chunk, embed, store and retrieve knowledge so agent responses can be grounded in relevant context.
Experiment with orchestration, parallel execution, supervision, conversation and debate-style coordination.
pip install agentic-ai-kit
The Python import package is:
agentic_ai
Agentic AI Kit currently uses Gemini-powered components for generation and agent execution.
Read the quick start
View the Python package
From practical agents to adaptive intelligence
Agentic AI Kit is designed for developers, researchers and data practitioners exploring questions such as:
How should an agent distinguish ordinary events from identity-shaping experiences?
How should corrected evidence alter derived memory state?
How can emotional relevance affect retention without overriding task relevance?
How can generative intelligence be introduced into existing SQL and analytical workflows?
How can agent systems remain modular rather than forcing every use case into one orchestration abstraction?
The project is open source and developed in public.
About the Author & Maintainer