Can I edit a video from Cursor?

Yes. With Valmera connected, Cursor's agent can upload a recording from your project, cut it, add captions, zooms and music, render a preview it checks itself, and export the final MP4 to a path in your repo.

Where does the Cursor mcp.json go?

Use ~/.cursor/mcp.json to make Valmera available in every project, or .cursor/mcp.json inside a project to scope it there. Add the valmera entry under mcpServers and keep your other servers.

Do I need an OAuth client ID or secret?

No. Valmera supports dynamic client registration with PKCE, so Cursor registers itself and opens the Valmera sign-in. A bearer token in headers is an alternative for environments without a browser.

Can Cursor upload a local video?

Yes. upload_start returns a presigned URL and the exact curl command; Cursor's agent runs it in the terminal, then calls upload_finish and waits for index_status. Files up to 100 GB and 3 hours in MP4, MOV, M4V, MKV or WebM.

Can Cursor export the final MP4?

Yes. The agent calls export_final for the version you approved, polls wait_for_job, then gets a link from download_url and can save the file with curl. The MP4 renders at full resolution from your original upload and ends with a brief Valmera end card.

Does connecting Valmera upload my repository?

No. Valmera only receives the tool arguments and files the agent sends it, such as the recording you ask it to upload.

Do I need a Valmera subscription?

Creating an account and uploading are free. Editing requires a subscription, and MCP access is included with eligible Valmera subscriptions; see plans at valmera.io/subscribe. Cursor is billed separately.

Video editing in Cursor with Valmera MCP

Connect Valmera and Cursor's agent can edit video without leaving your editor. It uploads a screen recording from your project, cuts the dead time, adds zooms, captions and music, renders a preview it checks itself, and exports the final MP4 into your repo. Valmera does the media processing in the cloud; nothing renders on your machine.

Add Valmera to Cursor

Add to Cursor (one click) → opens Cursor's install dialog with the entry below. Or paste it into ~/.cursor/mcp.json (every project) or .cursor/mcp.json (this project), keeping your other servers:

MCP.JSON · OAUTH

{
  "mcpServers": {
    "valmera": {
      "url": "https://valmera.io/mcp/server"
    }
  }
}

Cursor discovers Valmera's sign-in, registers itself and opens the Valmera login: there is no client ID, secret or token to paste. Sign in with the email or Google account you subscribed with and allow the connection.

No browser available, for example on a remote machine? Use a private token from the connect page and keep it in an environment variable:

MCP.JSON · BEARER TOKEN

{
  "mcpServers": {
    "valmera": {
      "url": "https://valmera.io/mcp/server",
      "headers": {
        "Authorization": "Bearer ${env:VALMERA_MCP_TOKEN}"
      }
    }
  }
}

Keep ${env:VALMERA_MCP_TOKEN} literal in the file and set the variable in the environment Cursor starts from. Never commit the token.

Edit your first video from Cursor

1
Connect Valmera
Use the one-click install or the mcp.json entry above, then sign in to Valmera and allow the connection.
2
Allow the read tools
Cursor asks before each MCP call. Most of an edit is reading (project_state, get_words, get_edl, look_at), so allow those in your Run Mode and keep approval on renders and exports.
3
Upload the recording
Ask the agent to create a project and upload a file from your repo. It calls upload_start, runs the returned curl command in the terminal, calls upload_finish and waits for index_status to report done.
4
Describe the edit
Say what to keep, cut and add. The agent edits a versioned edit decision list, so nothing touches your original file and any cut can be restored.
5
Review and export
The agent renders a preview and looks at its own frames. When you approve, it calls export_final, polls wait_for_job and downloads the MP4 with download_url.

A brief for a release-notes demo

Paste this into Cursor's agent and adjust the paths and instructions:

PROMPT

Use Valmera to edit ./recordings/feature-demo.mov for the release notes.
Create a project, upload the file and wait for the analysis to finish.

Keep the successful walkthrough. Cut the idle opening, the failed first
attempt and every pause longer than a second. Speed up page loads 2x,
zoom in on each click, enlarge the cursor and put the recording in a
screen frame. Add clean captions. Keep the 16:9 aspect ratio.

Render a preview, look at it, fix anything wrong, then export the final
MP4 and save it to ./dist/feature-demo.mp4.

For screen recordings the agent reaches for cut_range, set_speed, add_zoom, enhance_cursor, set_screen_frame and add_captions. It can also record a live site with record_website_demo, as long as the URL is publicly reachable: Valmera's browser can't see your localhost. All tools are in the tool reference.

Troubleshooting

Problem	Fix
Valmera is missing or shows no tools	Check that mcp.json is valid JSON and the URL is exactly https://valmera.io/mcp/server. MCP Logs in Cursor's Output panel show whether the connection or the sign-in failed.
“Connecting apps to Valmera is not enabled for this account”	Sign in with the Valmera account that has your subscription. MCP access is included with eligible subscriptions.
The bearer token isn't picked up	Cursor reads VALMERA_MCP_TOKEN from its own process environment. Set it where Cursor is launched, then restart Cursor and reconnect.
Every call asks for approval	Cursor asks before each MCP call by default. Allow Valmera's read tools (project_state, get_words, look_at, get_edl) in your Run Mode and keep approval on renders and exports.
A render says STILL RUNNING	Normal for slow work. The agent should call wait_for_job with the same job ID until it finishes, not start the render again.
Upload or analysis didn't finish	A file path alone isn't an upload. The agent must PUT the bytes to the presigned URL, call upload_finish and wait for index_status to report done.

Cursor's Run Modes documentation explains approvals and team policies.

Limits

Output is an H.264 MP4 with burned-in captions: no codec settings, no SRT or VTT, no project file for another editor.
Website recording needs a public URL, not your local dev server.
One editor per project at a time: Studio and Cursor can't edit the same project at once.

The full list is on the MCP server page.

Sources

Cursor configuration was checked against Cursor's documentation on September 9, 2026; Valmera details against its live server card on October 6, 2026.