Skip to content
PershotSTUDIO
Log in
Join the waitlist

Connect the AIyou already use

Pick your agent from the list. Each one has a setup page with three short steps and a link to its own docs. Once connected, it quotes first and makes videos from your creative balance, the one the studio uses, within the budget you set.

Real renders not open yet

Connecting your own AI agent is open to every account and needs no key. Developer API keys, for automating your own work from code, are by application.

Supported AI agentsone list for everyone

No paid placement: every agent is listed in the same order for everyone. Names are their makers' own, and each label says what we know.

What the labels mean

Tested by us
We connected it ourselves and ran a job from start to finish.
Documented
Its own docs describe connecting to a remote MCP server with sign-in, which is how Pershot works. It should connect; we have not tested it yet.
Needs a connection code
Its docs describe a header rather than sign-in, so it connects with a connection code you make under AI connections.
Has limits
It can work, with limits set by its maker: a plan, an admin setting, where it is offered or how it connects. Its setup page says which.

Three ways inone creative balance

Pick the one your agent already has. All three spend your creative balance, at the same prices as the studio. Developer API keys, for automating your own work from code, are by application and spend the separate developer API balance.

  1. 01 · Interface

    A link that fills the studio in

    For agents that browse like a person. One address carries the prompt and the settings; the studio opens filled in, with the price on the Generate button. A link never starts a render: someone presses Generate.

    Signing inIt works in your signed-in browser.

    Open the studio
    Address
    https://pershot.app/app?prompt=A+paper+boat+on+a+rain+puddle&model=mini&duration=8&resolution=480p&ratio=9:16

    prompt is required. Optional: model=fast|mini, duration=4 to 15, resolution=480p|720p, ratio=16:9|9:16|1:1, audio=0|1. For images: kind=image, tier=fast|mini, size=1K|2K.

  2. 02 · Command line

    A sign-in of its own, with nothing pasted

    For agents that run commands: a terminal, a server, a sandbox. The agent asks for a short code, you confirm it in your browser, and it gets a sign-in made for it. You choose what it may do and a monthly cap.

    Signing inDevice login: you confirm a code. No password or key is typed into the agent.

    About the command line
    Terminal
    # 1. Ask for a code. Show the link and the code to the person.
curl -X POST https://api.pershot.app/v1/device/code \
  -H "Content-Type: application/json" \
  -d '{"client_name": "my agent on build-server"}'

# 2. Every 5 seconds, ask whether they allowed it. "approved" carries the credential, once.
curl -X POST https://api.pershot.app/v1/device/token \
  -H "Content-Type: application/json" \
  -d '{"device_code": "dc_…"}'

    It needs a verified email only; connecting is free. The sign-in is listed under AI connections, where you can cap or revoke it, and it spends your creative balance, like the studio.

  3. 03 · MCP

    Tools inside Claude and Codex

    For AI agents that speak MCP. Connect the server once; the agent gets tools to quote, generate and fetch results, and asks before each paid call.

    Signing inOAuth in the client. Each connection can carry its own monthly cap.

    Set up MCP
    Terminal
    claude mcp add --transport http pershot https://mcp.pershot.app/mcp

    The same command works for any MCP client that takes a remote server address.

Whichever waythe same guardrails

Price first
Every job is quoted in credits and dollars before it runs. A job never costs more than its quote or its cap.
A cap per connection
Each connection and each developer API key can have a monthly cap. Without one, the balance it spends is the limit: the creative balance for a connection, the developer API balance for a key. Nothing is ever spent on credit.
Its own identity
Each agent gets its own connection, named on your AI connections page (a developer API key on your API keys page), so you can see what it spent and revoke it alone.
Safe retries
Each job carries an idempotency key. Sending the same request again returns the same job and never charges twice.
Failures
A render that makes no video (it fails or the model refuses it) or whose result a safety check blocks is refunded automatically, and the job says why. A video that is made is charged.
People stay in charge
A link never renders by itself, MCP clients ask before paid calls, and the agent skill has the agent confirm the quote with you first.

Before youhand it over

Does my agent need my password or my API key?

No. Through the interface it works in your signed-in browser. From the command line it asks for a short code that you confirm, and gets a sign-in of its own. Through MCP it signs in with OAuth. You never type a secret into the agent.

Can I limit what it spends?

Yes. Set a monthly cap when you allow it, or later under AI connections. Each job also carries the quote it was approved at and cannot cost more.

What can it not do?

It cannot see your payment details or change your plan. A connection can be limited to reading; a developer API key also to the models and tools you pick.

How do I take access away?

Revoke the connection under AI connections (a developer API key under API keys). It stops at once.