JsonCut
Start creating free
MCP FOR VISUAL AGENTS

Give your AI agent
a real creative workspace.

Create editable images and videos, bring in project media, review visible results and reuse approved templates from a compatible MCP client.

Connect JsonCut MCP Read the docsView an agent-built project
Images and videoVisual reviewTemplates and bulk
Create a launch video from these product files. Keep every scene editable.

I built four scenes and reviewed the opening before continuing.

Launch story00:32
PRODUCT UPDATETurn the work
into momentum.
4 editable scenesReviewed
jsoncut_review_scene
CONNECT JsonCut MCP

Give your agent
a visual workspace.

Choose your client for the right setup. Connect with your JsonCut account or use an API key in clients that support custom headers.

Connect ChatGPT

OAuth sign-in
  1. 01

    Add a custom MCP connection

    In ChatGPT on the web, open Settings → Security and login → Developer mode. Open Plugins, choose + and add JsonCut with the server URL below. Menu names and availability can vary by account and workspace policy.

  2. 02

    Connect with OAuth

    Choose OAuth, follow the JsonCut sign-in flow and select Allow access. Do not use a JsonCut API key as a client ID or client secret. Return to ChatGPT and enable JsonCut in your conversation.

  3. 03

    Check the connection

    Start a new conversation with JsonCut enabled and ask: “List my JsonCut projects.” Check that the result matches your Studio workspace.

Remote MCP server URL
https://mcp.jsoncut.com/mcp
THEN MAKE SOMETHING

“Create an editable launch image for my product and show me the result.”

THE TOOLS THAT MATTER

From first prompt
to a reviewed output.

The MCP surface stays focused on real production work. Internal Studio actions, billing and legacy JSON jobs are not exposed to new agents.

01

Create and edit

Start an image or video project, inspect its outline and make bounded changes without replacing everything.

jsoncut_create_projectjsoncut_edit_projectjsoncut_author_scene
02

Bring real media

Import a supported host file or public resource, transfer real bytes through an upload ticket, or continue with an upload in Studio when your client cannot send files.

jsoncut_import_host_filejsoncut_create_upload_ticketjsoncut_get_upload_handoff
03

Review the result

Render an image or a configurable video contact sheet so the agent can inspect composition, motion and timing.

jsoncut_validate_projectjsoncut_review_scenejsoncut_render_project
04

Reuse templates

Find a template, discover its typed fields and render one result or a complete batch with validated values.

jsoncut_list_templatesjsoncut_get_template_inputsjsoncut_bulk_render_template
See every MCP tool and recommended loop
FOR LONGER PROJECTS

Build scene by scene.
Keep each request small.

A short image or clip can be created directly. For a ten-minute video, the agent keeps one durable project, navigates its outline and works on one scene at a time.

  • Read the outline without loading the whole source
  • Patch one scene with local timing
  • Review frames before continuing
  • Render the pinned final project version
Campaign explainerPROJECT OUTLINE
01
Opening promise3 layers · reviewed
02
Product walkthrough8 layers · 2 media
03
Customer proof5 layers · reviewed
04
Final action4 layers · ready
OutlineAuthor sceneReviewContinue
author_scene(scene: "03", operations: […])Version 12
DESIGN ONCE · REUSE WITH CONTROL

Let the agent build the template.
Let data make the next 100.

Create the approved image or video once, expose only intentional inputs and reuse the same visual system from an Agent, a spreadsheet or a JSON array.

Agent-built template
WEEKLY PRODUCT UPDATEAcme Analytics

18.4% growth this week

Editable source
Typed inputs

companyText

growthNumber

logoImage

accentColor

Excel, CSV or JSON100 rows
Acme18.4%
Northstar12.1%
Atlas26.8%
100 reviewed variants

Discover before filling.The agent reads labels, types, required fields and examples instead of guessing from a thumbnail.

One contract for every path.The same inputs work for a single MCP render and bounded bulk production.

Keep the source editable.Open the template in Studio when the visual direction needs to change.

BEFORE YOU CONNECT

A focused surface.
No hidden handoff.

Does every project need scenes?+

No. Flat projects are the preferred path for simple images and short videos. Scenes become useful when a longer story needs incremental authoring and review.

Can an Agent upload local files?+

When the client can transfer real file bytes, it can use an authorized host file reference or a short-lived upload ticket. OAuth alone does not provide file access. If the client blocks uploads, the Agent can hand off to your Studio project, where you upload under Files and continue with the same media.

Can the Agent create freely?+

Yes. It can create editable text, media, shapes, SVG, groups and keyframes directly. Studio presets are optional inspiration, not a required authoring vocabulary.

What happens to existing V1 MCP clients?+

The legacy server remains available at its separate endpoint for compatibility. New clients should connect only to the V2 MCP server shown above.

ONE ENDPOINT · EDITABLE OUTPUTS

Connect the agent.
Keep the creative source.

Connect ChatGPT, Claude or another MCP client to build images and videos, upload resources, review frames and iterate on long projects safely.

Connect JsonCut MCP