Quickstart

Go from zero to your first video edit in under 5 minutes. Choose your path: AI assistant, CLI, or direct API.

Five minutes. That’s all it takes to import a video, run an AI edit, and see the result in Descript. Pick the path that matches how you work.

AI Assistant (fastest)

The MCP connector lets you edit video by talking to Claude, Cursor, or other AI assistants. No API token needed.

1. Connect Descript to Claude

   In Claude Desktop, click Customize > Connectors > find Descript > click Connect. Authorize with your Descript account.

   For Claude Code, run:

   claude mcp add descript --transport http https://api.descript.com/v2/mcp
   
2. Import a video

   Ask Claude:

   Create a Descript project from this video:
   https://test-files.descriptapi.com/demo-video.mp4
   

   Claude calls the API and creates the project for you.

3. Edit with AI

   Ask Claude:

   Add Studio Sound and captions to my latest project
   

   The agent applies the edits and tells you when it’s done.

4. Review in Descript

   Claude provides a project link. Open it in Descript to review the result and export.

Full MCP guide

Setup for Cursor, other agents, example prompts, and troubleshooting.

CLI

The CLI gives you interactive commands with built-in job polling. Good for testing and scripting.

1. Install the CLI

   npm install -g @descript/platform-cli@latest
   

   Requires Node.js 24+.

2. Configure your API key

   descript-api config set api-key
   

   Enter the API token from your Descript settings. See Authentication if you haven’t created one yet.

3. Import a video

   descript-api import \
     --name "My First Project" \
     --media "https://test-files.descriptapi.com/demo-video.mp4"
   

   The CLI polls for completion automatically and prints the project ID when done.

4. Edit with AI

   descript-api agent \
     --project-id YOUR_PROJECT_ID \
     --prompt "Remove filler words and add Studio Sound"
   
5. Review in Descript

   Open the project URL printed by the CLI to review and export in Descript.

Full CLI guide

All commands, profiles, file handling, and batch operations.

REST API

Direct HTTP access for building production workflows and integrations.

1. Get an API token

   Create a token in your Descript settings. See Authentication for details.

2. Import a video

   cURL

   curl -X POST https://descriptapi.com/v1/jobs/import/project_media \
     -H "Authorization: Bearer YOUR_API_TOKEN" \
     -H "Content-Type: application/json" \
     -d '{
       "project_name": "My First Video",
       "add_media": {
         "demo.mp4": {
           "url": "https://test-files.descriptapi.com/demo-video.mp4"
         }
       },
       "add_compositions": [
         {
           "name": "Demo Video",
           "clips": [{ "media": "demo.mp4" }]
         }
       ]
     }'
   

   Response:

   {
     "job_id": "project-media-import-9d635d5b",
     "drive_id": "c9c5c47e",
     "project_id": "e2f89ce6",
     "project_url": "https://web.descript.com/e2f89ce6"
   }
   
3. Wait for import to complete

   Poll the job status endpoint until job_state is "stopped":

   cURL

   curl https://descriptapi.com/v1/jobs/YOUR_JOB_ID \
     -H "Authorization: Bearer YOUR_API_TOKEN"
   

   Or pass callback_url in the import request to get a webhook notification instead.

4. Edit with AI

   cURL

   curl -X POST https://descriptapi.com/v1/jobs/agent \
     -H "Authorization: Bearer YOUR_API_TOKEN" \
     -H "Content-Type: application/json" \
     -d '{
       "project_id": "YOUR_PROJECT_ID",
       "prompt": "Remove filler words and add Studio Sound"
     }'
   
5. Check the result

   Poll the job status. When done, the result.agent_response field contains a summary of what the agent did:

   {
     "job_state": "stopped",
     "result": {
       "status": "success",
       "agent_response": "Done! I removed filler words and applied Studio Sound.",
       "project_changed": true,
       "ai_credits_used": 32
     }
   }
   

   Open the project_url in Descript to review and export.

API Reference

Full endpoint documentation with request/response schemas.

What just happened

In each path above, you:

1. Imported media — Descript created a project, downloaded the video, transcribed it, and set up a composition
2. Ran an agent edit — The AI agent analyzed the content and applied the requested changes
3. Got a project link — The finished project is ready to review and export in Descript

Both import and agent operations are asynchronous jobs. Learn more about how jobs work in Working with Jobs.

Next steps

• Common Workflows — Patterns for batch processing, webhook callbacks, and multi-step edits
• API Reference — Full endpoint documentation
• Known Issues — Current limitations and workarounds