VidbyteVidbyte

Browse docs

Run your first agent

APIDashboard
OverviewQuickstartAPI KeysPricingSDKsCLIQuickstartRun your first agentAgent skillsPlatform connections

After the quickstart

Run your first agent

With vidbyte-cli installed and signed in, pick the agent that fits your job, finish the one setup step it needs, and run it. Every agent also has its own page on vidbyte.pro with use cases, examples, and a setup prompt.

Have these ready

The quickstart covers all of this. Check each item once, then pick an agent.

  • A signed-in CLI — vidbyte-cli whoami prints your account. If it doesn't, finish the quickstart first.

  • Balance on your account — Every agent run is paid from your Vidbyte API balance, and a run is refused rather than half-finished when the balance is empty.

  • Codex for local agents — Task Board, Persistent Agent, Ensemble Agent, and Stages Agent drive Codex on your computer, with model usage billed to your own OpenAI account.

Open the CLI quickstart

Check your API balance

Pick an agent for the job

Each agent does one kind of job. Start with the one whose job matches yours; you can run the others later with the same account and balance.

Suggestion Agent
Use it whenYou want the next best moves toward a goal, ranked.
Where it runsStarted from your terminal; Vidbyte runs the model calls
Vidbyte chargeUsage-based: each model call at its token rate plus 2%
Persistent Researcher
Use it whenA question needs sourced research you can come back to.
Where it runsVidbyte's servers
Vidbyte chargeUsage-based
Task Board
Use it whenYou have a list of tasks, each deserving a fresh agent.
Where it runsCodex on your computer
Vidbyte charge$0.02 per run
Persistent Agent
Use it whenOne coding task where agents stop too early.
Where it runsCodex on your computer
Vidbyte charge$0.02 per launch
Ensemble Agent
Use it whenYou want several approaches compared before coding.
Where it runsCodex on your computer
Vidbyte charge$0.02 per launch
Stages Agent
Use it whenOne job runs through phases with their own instructions.
Where it runsCodex on your computer
Vidbyte charge$0.01 per run
Adversarial Team
Use it whenYou want hostile reviewers to attack a change. Coming soon.
Where it runsCodex on your computer
Vidbyte charge$0.25 per launch

For agents that run on Codex, the Vidbyte charge is only part of the cost: the model work is billed to your own OpenAI account. vidbyte-cli runtime list prints the current price of each local agent.

Browse every agent on vidbyte.pro

Set up and run the agent you picked

Open the tab for your agent. Each one lists its one setup step, the command that runs it, and how to follow up on the result.

  1. Use a key that can call models

    Each model call is billed to your balance, so your Vidbyte key needs the models:invoke permission; admin and full-access keys also work. No Codex or OpenAI key is needed.

    Manage API keys

  2. Ask for your next actions

    Describe the goal, what is done, and what is in progress. The agent returns ranked next actions after a critic reviews each idea.

    Next actions

    bash
    vidbyte-cli agents suggest run --goal "Ship the billing page this week" --completed "Checkout flow works end to end" --in-progress "Invoice PDF rendering" --horizon now --count 3

  3. Save the ideas as JSON

    The root --format json option writes the result as JSON, so you or your coding agent can pick an idea by its id.

    Save the result

    bash
    vidbyte-cli --format json agents suggest run --goal "Ship the billing page this week" --count 5 > suggestions.json

  4. Hand one idea to your coding agent

    Turns one saved idea into a handoff packet a coding agent can start work from. The Suggestion Agent itself never edits files.

    Handoff

    bash
    vidbyte-cli agents suggest handoff --input suggestions.json --idea idea-001
    Open the Suggestion Agent page

Open the Adversarial Team page

Options for every run

These root options and checks work with every agent above. Root options always go before the command name.

Get machine-readable output

--format json switches results to JSON records for scripts and coding agents. Results go to stdout, while progress and errors stay on stderr.

JSON output

bash
vidbyte-cli --format json research threads
Use a separate profile

--profile keeps a separate set of credentials and settings under a name, which keeps work and personal accounts apart.

Profiles

bash
vidbyte-cli --profile work research threads
See every local agent and its price

Fetches the current catalog of local agents with the price and supported host for each, so you know what a launch costs before you start one.

Catalog

bash
vidbyte-cli runtime list

Next steps

Agent skills

Paste one prompt into your coding agent to put these agents to work on a real problem.

Platform connections

Let the CLI read GitHub, Slack, or Google Drive context.

All agents

Every agent's use cases, examples, and setup prompt.

Pricing

How agents charge and how to add usage to your account.