About Tobub
Understand Tobub’s working model, core objects, and the tasks it is built to carry through.
What Tobub is for
Tobub is a local-first workspace built around AI digital employees. You give it an outcome, not only a question. A digital employee breaks the goal into steps, selects the right Skills, runs them through the Runtime, and keeps the process, evidence, and result together.
It fits research, monitoring, analysis, content planning, data preparation, and repeatable operations. For example:
Review today’s industry news, verify the sources, and prepare the three actions worth taking next.
This kind of work usually crosses several pages and sources, requires information to be organized and checked, and ends with a result someone can act on. Tobub turns those steps into a workflow that can be inspected, reviewed, and continued.
Five core objects
- Agent / digital employee: has a role, goal, and work plan, and coordinates the task.
- Skill: a JSON-defined executable capability with declared parameters and SDK access to browser, platform, or Runtime resources.
- Runtime: the local execution layer for browsers, files, accounts, assets, and Sessions.
- Platform / Rule: a Platform describes the target system, URLs, domains, and capabilities; Rules keep page interactions stable and reusable.
- Session: the record of one piece of work, including the plan, progress, sources, evidence, result, and context.
The short version is: the Agent decides what to do, a Skill defines how to do it, the Runtime provides the environment, Platform/Rule connects the page, and the Session records what happened.
How AI exploration leads to Skills
When you know the outcome but not the page structure or a reliable interaction sequence, use AI exploration in /discover. It observes pages in a controlled browser, records actions and evidence, and produces page knowledge plus a reviewable Skill proposal.
The proposal is not an automatically saved finished Skill. Review its parameters and side effects, complete the code, and validate it before saving. Once verified, the Skill can run directly or be composed by an Agent.
Tobub and this repository
This repository contains open Skill definitions, documentation, and examples. It does not contain the Tobub desktop application, the Runtime implementation, platform authentication, or the complete execution environment. Running a Skill requires compatible Tobub software or Runtime.
A sensible first-use path
- Install Tobub and confirm that the Runtime is available.
- On the first Home visit, run the onboarding demo and open the Session it creates.
- Check the system settings for timezone, data location, and required Runtime connection details.
- Configure a Provider, model, and API key in AI settings.
- Use
/discoverfor a read-only target to verify the browser and evidence path. - Review, create, or import a Skill in
/skill, then hand it to an Agent or schedule it for repeatable work.