Tinybot
A desktop AI agent works with project files through a connected model, with reusable workflows, scheduled tasks and MIT-licensed source code.

Tinybot is a desktop AI agent for people who want an assistant to work with their files and project folders. Workspaces keep related conversations together, and execution records let you inspect what the agent did and what it produced.
- File review: Preview output files, add annotations and compare versions inside the app. Browser annotation supports feedback on web content, and the agent can generate custom charts.
- Reusable workflows: Agent Graphs let you create, edit and repeat workflows. A conversation agent can call a saved graph from its selected workspace and use the result to continue its task.
- Agent teams: Teams divide work around a shared goal, with plans, handoffs and results available for review.
- Recurring tasks and memory: Scheduled tasks produce reports and support manual reruns. Long-term memory stores preferences and project facts for later work.
- Extensions: Plugins, Skills and MCP connections add capabilities. A built-in helper converts existing Skill or MCP configurations into Agent Plugins, and Hooks support scripts at task stages.
Tinybot is open source under the MIT license and uses a Rust-based desktop app. It requires a connected model provider. Model usage statistics distinguish cached tokens, reasoning, background calls and retries.
On Windows, an animated desktop pet provides quick chat with text, images or local file attachments. The same conversation can continue in the main window.