Flagship Desktop AI Assistant

Autonomous Intelligence.
Integrated in Desktop.

Agentic is a powerful, local-first desktop AI assistant built on the Pytron framework. It combines state-of-the-art LLM reasoning with a massive library of 70+ tools, deep system integration, and a stunning modern UI.

Download Latest Explore Pytron-Kit

Transparency & Strategy

The Architecture of Openness

A common question we encounter: "If Agentic is hosted on GitHub, why is the application code private?"

We believe in radical openness where it matters most. Agentic is delivered as a high-performance, strategic binary to ensure a secure and consistent flagship experience. However, the "magic" isn't a secret.

Every core system—from the Python-to-Web bridge to the multi-agent orchestration—is powered by the Pytron-Kit.

The Experience Agentic is the Product.

The Infrastructure Pytron-Kit is the Engine.

If you are here to audit our logic or build your own elite AI applications, the Pytron-Kit repository is where the open-source heart of this project lives. We aren't just building a tool; we're building the framework for the next generation of desktop intelligence.

Architecture & Stack

Layer	Technologies
LLM Orchestration	LangChain, Ollama, Google Generative AI
Memory / Vector DB	Faiss-CPU, FastEmbed
Web Automation	Playwright, BeautifulSoup4
Backend Core	Python 3.10+, Pytron-Kit
Frontend Core	React 18+, Vite, JavaScript
UI/UX	Vanilla CSS (Glassmorphism), Lucide Icons

Infinite Capabilities

A multidimensional Approach to autonomous work powered by pytron-kit

Core Engine

Agentic Reasoning

Plan, execute, and reflect on complex multi-step tasks autonomously. Our recursive reasoning engine handles the depth, so you can focus on the goal.

70+ Native Tools

A massive library of built-in capabilities including Playwright for web scraping, GitHub for dev-ops, and sandboxed Python for data science.

Long-Term Memory

Local vector storage (RAG) using FAISS. The agent learns your style, project context, and preferences over time—completely offline.

Artifact Panel

Live previews for code, diagrams, and data visuals.

Multimodal Vision

Analyze images, PDFs, and screen content instantly.

Plugin System

Easily extend the engine by dropping Python scripts into the plugins directory. Automatically detected and ready to call.

MCP Support

Connect to the global Model Context Protocol ecosystem.

Voice Actions

Dictate commands and receive high-fidelity audio feedback.

The Toolbelt

Granular control over your digital environment.

PlaywrightWikipedia SearchDuckDuckGoArticle FetcherYouTube TranscriptProxy Rotator Python 3.11 SandboxGit/GitHub APIMermaid JSESLintPrettierProject TreeShell Access Vector RAGPDF OCRDALL-E 3Audio STT/TTSFAISS Vector DBCSV Analysis Process ManagerResource MonitorClipboard SyncKeyring AuthFile WatcherNotifications PlaywrightWikipedia SearchDuckDuckGoArticle FetcherYouTube TranscriptProxy Rotator Python 3.11 SandboxGit/GitHub APIMermaid JSESLintPrettierProject TreeShell Access Vector RAGPDF OCRDALL-E 3Audio STT/TTSFAISS Vector DBCSV Analysis Process ManagerResource MonitorClipboard SyncKeyring AuthFile WatcherNotifications

User Guide & Documentation

Everything you need to master Agentic and build the future of your workflow.

Quick Start

Getting a Gemini API Key

To use Google's powerful cloud models:

Go to the Google AI Studio.
Sign in and click "Create API key".
In Agentic, go to Settings -> System -> Google Gemini API and paste your key.

Installing Ollama (Local Models)

Download from ollama.com.
Run the installer and ensure the icon appears in your tray.
Pull models via terminal (ollama run llama3) or in Settings -> System.

App Customization

Settings & Notifications

Press Ctrl + , to open settings. You can manage notifications, default models, and tool permissions here.

Theme & Colors

In Settings -> General, use the circular picker to change your Primary Accent Color. Click Save to see the glow update instantly across the entire interface.

Custom Plugins

How to add Plugins

Agentic is highly extensible. You can add your own Python tools by dropping them into the plugins folder:

                                        Windows: %APPDATA%/Agentic/plugins

                                        macOS: ~/Library/Application Support/Agentic/plugins

Stricter Gemini Integration

To support stricter tool-calling rules, use the @register_tool decorator and Pydantic models for input validation.

                                        # plugin_file.py

                                        from pydantic import BaseModel, Field

                                        from pytron.tools import register_tool

                                        class MyInput(BaseModel):

                                            topic: str = Field(..., description="Topic to analyze")

                                        @register_tool(description="Analyzes a specific topic using local tools")

                                        def analyze_topic(data: MyInput):

                                            return {"status": "Success", "data": f"Analyzed {data.topic}"}

Security & Responsibility

Virus Warning: Adding custom scripts involves executing code on your machine. Never add plugins from sources you don't trust.

Disclaimer: We are NOT responsible for any system damage or data loss caused by third-party plugins.

Modes & Commands

@web Internet Search

@code Coding Assistant

@reasoning Multi-step Planning

@file Local File Analysis

Common commands: /clear, /reset, /history, /help.

FAQ

Is it free?

The standard version is free. We believe in open access to AI.

Data Privacy?

100% local. Your vector database stays on your disk.

Latest Updates

Fetching latest release info...