Introducing Vidbyte Agents

A general assistant can help with almost anything. A specialized agent starts from a particular kind of work: it follows a workflow shaped for that job and returns something you can use, such as a sourced research thread, a ranked set of next steps, or changes made in your project.

Vidbyte's agents are not one model wearing five names. They are different ways to move work forward, each with its own scope, setup, and output. Choose by the work in front of you, then open that agent's page for the exact instructions and current price.

Start with the kind of help you need. The five available agents below cover work that benefits from durable evidence, proactive ideas, continued effort, alternative approaches, or a clear sequence of tasks.

• Need evidence you can return to and extend? Start a research thread with Persistent Researcher.
• Unsure what would make the most difference next? Ask Suggestion Agent to find and rank possible actions.
• Know the task, but want it worked through beyond one coding-agent response? Use Persistent Agent.
• Still weighing how to solve a problem? Let Ensemble Agent compare approaches before implementation.
• Already have a trusted list of steps? Give Task Board the plan and its rules for context and failure.

These are different work patterns, not interchangeable levels of autonomy. Each agent page explains what it will and will not do, so you can match the run to the decision or work you want to delegate.

Persistent Researcher runs on Vidbyte's servers, so the research can continue without your computer staying on. Its usage-based price is tied to the work a run performs.

The other four agents run locally through Codex and use your own OpenAI account for model work. Vidbyte charges a separate flat admission fee for each run or launch; provider model usage is billed separately. Check an agent's page for its current setup and price before you start.

You do not need a new chat product to try an agent. Pick a job, read its page, and follow the setup prompt using the coding agent or assistant you already use. The agent pages include their own commands and explain what to check before a run is charged.