What is Tokeny?

Tokeny is a desktop AI assistant designed to turn a user's one-sentence instruction into complete deliverables such as PPTs, spreadsheets, reports, and documents. It uses a multi-agent system where a main Agent breaks down tasks and delegates them to specialized sub-agents for collaborative completion. The product supports features such as a skill marketplace, MCP tool integration, persistent memory, asset library, and browser operations, and allows users to connect their own API keys to use multiple model channels.

How to use Tokeny?

Download and install the Tokeny desktop client (supports macOS, Windows, Linux). After launching, describe your needs in natural language in the chat interface, for example, "Help me turn the quarterly summary into a 15-page presentation PPT." Tokeny's main Agent will automatically break down the task, call the appropriate skills or sub-agents to execute it, and save the final deliverable file locally. Users can also install more skills through the skill marketplace or connect their own API keys to use different models.

Core features of Tokeny

Agent Chat: Generate finished files such as PPT, Excel, reports, and Word documents through natural language instructions
Multi-Agent Collaboration: Main Agent breaks down tasks, specialized sub-agents execute in parallel, supports group chat collaboration and subtask delegation
Skills & MCP Marketplace: 14 built-in skills, supports installing community skills and integrating MCP standard tools
Persistent Memory System: 7 types of memory carry across sessions, viewable and editable, data stored only locally
Asset Library: Automatically accumulates deliverables such as reports, presentations, and notes, supports full-text search and task references
Browser Operations: Takes over logged-in browsers, automatically fills forms, scrapes, and submits, with screenshots saved throughout for evidence
Voice Input: Chinese voice-to-text, automatically polished into clean instructions
Model Channels: Supports 10 model channels and 60+ models, with customizable channels compatible with OpenAI/Anthropic protocols