Developers
Vecta Studio MCP server
Let Claude, Cursor or your own agent make videos with Vecta Studio: plan, script, cast, film, fix and publish, using your workspace's credits and approvals.
Get an API token01Overview
Vecta Studio runs an MCP server (Model Context Protocol), so AI agents can use Vecta the way a person does in the app: start a project, write the script, create the cast, film, fix shots and publish. Every step uses the same prices, credits and checks as the web app.
It is a remote server over streamable HTTP. Agents sign in with a Vecta API token sent as a bearer token.
Server URL
https://avs-prod-backend-937508501188.us-central1.run.app/mcp/Authentication
Authorization: Bearer vecta_pat_…02Get an API Token
Workspace admins create tokens in the app. Open Connect Claude (sign in first if asked, then open the link again) to go straight to your workspace's API tokens for agents card.
- 1. Name it after the agent that will use it, e.g. “Claude Code” or “Autopilot”.
- 2. Choose who it acts as: yourself, another member, or New agent member for an automated agent.
- 3. Set a monthly credit limit. Paid steps that would go past it are refused.
- 4. Copy the token. It is shown once. The card also gives you a ready-made Claude Code command.
03Connect Your Agent
Add the server to your agent with your token in place of vecta_pat_….
claude mcp add --transport http vecta https://avs-prod-backend-937508501188.us-central1.run.app/mcp/ \
--header "Authorization: Bearer vecta_pat_…"{
"mcpServers": {
"vecta": {
"url": "https://avs-prod-backend-937508501188.us-central1.run.app/mcp/",
"headers": { "Authorization": "Bearer vecta_pat_…" }
}
}
}import anthropic
client = anthropic.Anthropic()
response = client.beta.messages.create(
model="claude-opus-5-5",
max_tokens=4096,
betas=["mcp-client-2025-11-20"],
mcp_servers=[{
"type": "url",
"url": "https://avs-prod-backend-937508501188.us-central1.run.app/mcp/",
"name": "vecta",
"authorization_token": "vecta_pat_…",
}],
tools=[{"type": "mcp_toolset", "mcp_server_name": "vecta"}],
messages=[{"role": "user", "content": "List my Vecta projects."}],
){
"mcpServers": {
"vecta": {
"command": "npx",
"args": ["mcp-remote", "https://avs-prod-backend-937508501188.us-central1.run.app/mcp/", "--header", "Authorization:${AUTH_HEADER}"],
"env": { "AUTH_HEADER": "Bearer vecta_pat_…" }
}
}
}claude.ai custom connectors
Custom connectors on claude.ai sign in with OAuth, which the Vecta MCP server doesn't support yet. Use Claude Code, Claude Desktop with mcp-remote, Cursor or the Claude API meanwhile.
04Making a Video
A video is a project that moves through these steps. Paid steps return straight away and keep running in Vecta, so agents check get_project instead of waiting, and never start the same paid step twice.
- 1.
capabilities— Learn the styles, video types, lengths, voices and models, and what Vecta can't do yet. - 2.
create_project— Free. Describe the idea and choose the style, aspect ratio, length and voices. - 3.
write_script— Paid. Writes the script and plans the shots. Read the result with get_project. - 4.
create_looks— Paid. Creates the cast and place images that keep faces consistent from shot to shot. - 5.
film— Paid. Films every shot and assembles the video. Poll get_project every 30–60 seconds; when video.url is set, it's ready. - 6.
get_video / publish_video— Get the MP4, or publish it to a connected YouTube or Instagram channel once a person has approved it.
Fixing a finished video
For a small visual change, use
edit_take. For a new action, framing or words, use edit_shot then refilm_shot. Then call reassemble to put the new take into the video. ask_vecta explains what went wrong in a shot, for free.05Tools
Read tools only look. Free tools change things without spending credits. Paid tools spend credits and accept max_credits.
| Tool | Cost | What it does |
|---|---|---|
| capabilities | Read | What Vecta can make: steps, styles, video types, voices, lengths, video models, and what it can't do yet. Call it first. |
| get_credits | Read | The workspace's credit balance and how much this token has used this month. |
| list_projects | Read | Recent video projects in the workspace, newest first. |
| create_project | Free | Start a project from an idea: style, video type, aspect ratio (9:16, 16:9, 1:1), length, voices, sound, language. |
| update_plan | Free | Change the plan of a project (idea, style, length, voices, model, title). |
| get_project | Read | Everything about a project now: stage, script, shots, cast looks, running jobs, the finished video and the next step. |
| quote | Read | Estimated credits for a paid step (script, looks, film, refilm, edit_take, reassemble). |
| write_script | Paid | Write the script and plan the shots (about a minute). With notes, rewrites them. |
| create_looks | Paid | Create the images of the cast and places, which keep faces consistent across shots (1–2 minutes). |
| film | Paid | Film every shot, then assemble the video with transitions and narration (usually 5–15 minutes). |
| edit_shot | Free | Change one shot's description, narration, dialogue, length, sound or starting image. |
| refilm_shot | Paid | Film one shot again from its current description. |
| edit_take | Paid | Change something in a filmed shot and keep the rest, e.g. “remove the microphone” (9:16 and 16:9). |
| reassemble | Paid | Put the shots together again after a re-film or edit (small cost). |
| ask_vecta | Read | Ask Vecta's own assistant about a project in plain language; it explains and suggests, it doesn't change anything. |
| get_video | Read | The finished video: MP4 link, length and every finished version. |
| list_channels | Read | Publishing channels connected in the workspace (YouTube, Instagram). |
| publish_video | Paid | Publish the finished video with a title, caption, tags and privacy. Only after a person approved it. |
| publish_status | Read | Where the video was published, with links and any errors. |
| request_feature | Free | Tell the Vecta team about something a video needs that Vecta can't do yet. |
06Credits & Limits
Workspace credits
Paid steps spend the workspace's credits, at the same prices as the app.
get_credits shows the balance.Quote first
quote returns the credits a step will cost. Pass max_credits to a paid tool and it refuses anything pricier, starting nothing.Monthly token limit
Each token can have a monthly credit limit. Steps that would pass it are refused; the limit resets on the 1st (UTC).
07Errors
Problems come back as tool errors in plain language, saying what to do next.
401 Unauthorized
The Authorization header is missing, or the token is wrong, revoked or expired. Create a new token.
Over the token's monthly limit
The paid step would pass the token's monthly credit limit. Nothing was started. Raise the limit or wait for the next month.
Over max_credits
The quote is above the max_credits you passed. Nothing was started.
Out of credits
The workspace has no credits left for this step. Add credits in Vecta, then try again.
Step out of order
For example “Write the script first”. Errors say what to do next; follow get_project's next_step.
08Security
A token is a member
It acts as one member of one workspace, with that member's permissions. It can't manage tokens or switch workspaces.
Agent members can't sign in
Choose “New agent member” when you create a token for an AI agent: Vecta adds a member with no email or password, so the token is its only way in.
Shown once, stored hashed
Vecta keeps only a hash. Copy the token when it's created and keep it in a secret manager or environment variable, never in code.
Revoke anytime
Revoking a token in the workspace's API tokens card stops it immediately.
09Example Prompts
Once connected, just ask your agent in plain words:
- “Make a 30-second vertical realistic video about the last lighthouse keeper in Cornwall. Quote filming first and don't spend more than 3,000 credits.”
- “Show me my last three Vecta projects and tell me which ones still need filming.”
- “In my latest project, shot 3 has an extra person. Fix it with the cheapest option and reassemble the video.”
- “Write a title and description for my finished video and publish it to YouTube as unlisted after I say yes.”
Questions? Contact us.
Back to top ↑