New Types of Agents

A Vidbyte agent is a specialized agent that does one job well and returns evidence you can check: filed sources, opened diffs, a drawn plan, a listed set of objections. Each agent page on this site names its use cases, its when-to-use rules, and a setup prompt you copy into the coding agent or assistant you already use.

Vidbyte's lineup combines three traits: specialized agents focus on a defined job and return its useful artifact; proactive agents can keep work moving after your first instruction; autonomous agents carry a task through their defined steps and budget. An individual agent page explains which traits apply and what that agent actually does.

A general assistant responds to many kinds of requests, often with an answer to interpret or turn into action. A Vidbyte agent is configured around a particular job: it follows that job's workflow, works within its stated boundaries, and returns a concrete result such as filed sources, a plan, or a code diff. Check the individual agent page for its exact scope and output.

Yes. You can paste an agent's setup prompt into the coding agent or assistant you already use, including Codex or Claude Code, and have it guide the setup. That prompt is the onboarding path; the Vidbyte agent itself runs on its listed surface. For local runtimes, supported coding-agent hosts vary by runtime, so check that agent's page before launching it.

Most agent runs are charged against your prepaid balance. Hosted agents are usage-based: metered as the work runs, so the final cost varies with what the run does. The Suggestion Agent is billed the same way, per model call, at that model's per-token rate plus 2%. Local runtimes have a flat Vidbyte admission charge per launch; some also let you pay that charge directly through x402. Each agent page and the runtime list show the applicable price.

You pay Vidbyte for the agent run, for the model calls the Suggestion Agent makes through Vidbyte, or for a local-runtime admission. For local runtimes that drive a coding agent on your own machine, model usage is separate and billed to your own provider account. The Suggestion Agent has no separate provider bill, because its model usage is what Vidbyte meters. Check the agent page for its Vidbyte charge and your provider's pricing for model costs.

• Card: Buy prepaid USD balance on your account at the usage page. Hosted runs, Suggestion Agent model calls, and local runtime launches can charge against that balance.
• x402: Pay a local runtime's launch admission directly with a compatible x402 wallet when that runtime supports it; no prepaid balance is used for that launch. Suggestion Agent model calls are not paid this way and always draw from the prepaid balance.
• MPP: A compatible agent client can use MPP to add funds to your Vidbyte API balance through the agent top-up route; later balance-funded runs draw from that balance.

Yes. One Vidbyte API key is required: it identifies your account, owns what your agents create, and ties charges to you. Sign in with Google, create a key at https://vidbyte.pro/settings/api, and save it securely because it is shown only once. If you lose it, create a replacement there and update your saved credentials; the old plaintext cannot be shown again.

Start with the agent marketplace and choose the task you need done. Each agent page explains its use cases, when to use it, where it runs, and how it is charged. Copy that page's setup prompt into the coding agent or assistant you already use, then follow its checks and setup steps before launching.