MCP
Use StoryCapture from an AI client
The StoryCapture desktop app runs an MCP server on your computer. Connect Claude or another MCP client and it can plan a demo video with you, write and rehearse the story against the real page, record it in the app, review the export and publish it.
Last updated
1. Turn on the server
In the StoryCapture app open Settings → Integrations and turn on MCP server. It listens only on this computer, on port 47823 unless you change it, and answers only clients that send its token, shown on the same page. New token makes a new one; clients set up with the old one stop working.
2. Connect your client
Claude Code
Copy the command from Integrations, or run it with your token:
claude mcp add --transport http storycapture http://127.0.0.1:47823/mcp --header "Authorization: Bearer <token>"Claude Desktop
Install the StoryCapture extension from Claude's extension directory, or open the storycapture.mcpb file, and paste the token when it asks. The token is kept in your system keychain.
Cursor, VS Code, Windsurf and other clients
Clients that start servers over stdio run the @storycapture/mcp bridge with Node.js 20 or later. It relays your client to the app on this computer and nowhere else. Add it to the client's MCP settings:
{
"mcpServers": {
"storycapture": {
"command": "npx",
"args": ["-y", "@storycapture/mcp"],
"env": { "STORYCAPTURE_MCP_TOKEN": "<token from Settings → Integrations>" }
}
}
}Set STORYCAPTURE_MCP_PORT too if you changed the port. A token in a config file is kept in plain text there; New token in Integrations revokes it.
3. Make a video
Ask your client for a demo video of a page, or start the write_demo_video prompt. It interviews you, writes a beat sheet for you to approve, then builds and rehearses the story before anything is recorded. For a video already published, the improve_published_video prompt reads where viewers leave and re-records that one scene.
What asks you first
These tools wait for you to confirm in the app, unless you turn on Full access in Settings → Integrations:
record: Recordupload: Upload a videoresume_upload: Resume an uploadset_playing_version: Set the playing version
A story of a website records only its own browser window, not your screen. Uploads go to your StoryCapture workspace. See the privacy policy for what an AI client receives.
Tools
Plan the video
video_script_guide: Read the video guidebeat_template: Beat sheet templatevalidate_beat_sheet: Validate a beat sheetstory_skeleton: Story skeletondsl_reference: Story language referencelist_card_styles: List card stylespreview_card_style: Preview a card stylewrite_card_template: Write a card templatedelete_card_template: Delete a card templatelist_sounds: List sounds
Write and rehearse the story
list_projects: List projectscreate_project: Create a projectread_story: Read a storywrite_story: Write a storyvalidate_story: Validate a storyinspect_page: Inspect a pagerun_simulator: Rehearse in the simulator
Record, review and export
record: Recordget_recording_status: Recording statuscancel_recording: Cancel a recordinglist_recordings: List recordingscheck_take: Check a takepreview_look: Preview the lookexport: Export a videoget_job_status: Export statuslist_exports: List exportsreview_export: Review an export
Publish and improve
list_workspaces: List workspacesupload: Upload a videoget_upload_status: Upload statuslist_pending_uploads: List unfinished uploadsresume_upload: Resume an uploadwatch_analytics: Watch analyticsset_playing_version: Set the playing version
Troubleshooting
- "StoryCapture is not answering on port …": open the app and turn on the MCP server in Settings → Integrations; check the port matches.
- "StoryCapture refused the token": a new token was made; copy it again from Integrations into your client.
- A recording or upload never starts: the app is waiting for your confirmation; bring its window forward.