agentleFS
Sign inStart free
descriptincdescriptinc/descript-mcp/AGENTS.md

descript-mcp

AGENTS.md37 starsChanged 43 days agoApache-2.0View on GitHub

Best practices for working with the Descript MCP tools.

  • Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.

    Of the AGENTS.md files we read, the 31,774 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.

  • In this repository5 other files it gives its agents.

The file

We split this AGENTS.md into the rules it gives an agent

9 rules. About 0 are common, 1 is distinctive, and 8 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 8

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use search_drive to find projects, folders, media, layout packs, and transcript matches across the connected Drive (by name or …

    line 5
  2. Very specific:

    Use get_project to inspect a project's compositions and media files before making edits.

    line 6
  3. Very specific:

    When creating a new project with media, include add_compositions so imported media appears on the timeline.

    line 7
  4. Very specific:

    When importing into an existing project, omit add_compositions unless explicitly asked, to avoid disrupting existing edits.

    line 8
  5. Very specific:

    Use prompt_project_agent for any editing task: trimming, rearranging, removing filler words, adding captions, generating AI images/video, …

    line 12
  6. Very specific:

    If a project has more than one composition, pass composition_id explicitly to prompt_project_agent, export_transcript, and publish_project …

    line 13
  7. Very specific:

    Use get_project after edits to verify the result and retrieve composition IDs for publishing.

    line 14
  8. Very specific:

    Wait by looping wait_for_job with wait_seconds: 20 (set it explicitly — don't rely on the default) until the job reaches a terminal state.

    line 18
  9. Distinctive:

    Never start a second agent job for the same clip while one is still in flight.

    line 20

≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
AGENTS.md

Descript MCP workflows

Best practices for working with the Descript MCP tools.

  • Use search_drive to find projects, folders, media, layout packs, and transcript matches across the connected Drive (by name or spoken/content text). Use list_projects to paginate or filter the project list (folder, dates, creator). Never guess or fabricate project IDs.
  • Use get_project to inspect a project's compositions and media files before making edits.
  • When creating a new project with media, include add_compositions so imported media appears on the timeline.
  • When importing into an existing project, omit add_compositions unless explicitly asked, to avoid disrupting existing edits.
  • import_media, prompt_project_agent, and publish_project return immediately with a job_id. Always follow up with wait_for_job — see "Waiting on async jobs" below for the polling loop.
  • For direct file uploads, use content_type and file_size instead of url. PUT the file to the returned upload_url with header Content-Type: application/octet-stream.
  • Pass media URLs as-is to import_media — do not rewrite Google Drive or Dropbox share links.
  • Use prompt_project_agent for any editing task: trimming, rearranging, removing filler words, adding captions, generating AI images/video, importing stock media, and more. Describe the desired edit in natural language.
  • If a project has more than one composition, pass composition_id explicitly to prompt_project_agent, export_transcript, and publish_project — do not rely on the default.
  • Use get_project after edits to verify the result and retrieve composition IDs for publishing.

Waiting on async jobs

  • Wait by looping wait_for_job with wait_seconds: 20 (set it explicitly — don't rely on the default) until the job reaches a terminal state. Surface progress.label and percent after each poll so the user sees the job advancing.
  • A timeout error (code -32001) means the job is still running, not failed. Poll the same job_id again — never re-issue the original request in response to a timeout.
  • Never start a second agent job for the same clip while one is still in flight.
  • wait_seconds: 0 returns the current status immediately — use it for a plain status check, not to wait.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free