YouTube API quota exceeded: what it means and how to fix it

A quotaExceeded (403) error means your Google Cloud project has used up its daily YouTube Data API allowance, and YouTube will refuse further requests from it until the quota resets¹. Uploads have a bucket of their own, so the error can hit after 100 videos.insert calls even when the rest of your quota looks untouched². The fastest way out is to stop uploading through your own project: connect the channel to PublishQ and send the video through its API, CLI or MCP server. You don't need a Google project, so there is no quota of yours to run out.

What Google gives every project by default

Google's quota page lists the default allocation for any project that turns on the YouTube Data API²:

A few more rows from the same table matter for an upload pipeline. Setting a custom thumbnail with thumbnails.set costs 50 units from the shared 10,000, and so do videos.update, playlistItems.insert and captions.delete. Uploading a caption track with captions.insert costs 400². Daily quotas reset at midnight Pacific Time².

You can watch your usage on the Quotas page of the Google Cloud console². If that page shows the uploads bucket at 100 of 100, you have your answer.

Why it runs out sooner than you expected

The allowance belongs to the project, and the project is shared by every script, cron job and user that authenticates through it. A few patterns burn through it quickly:

• Retry loops. Every request costs at least one unit, invalid ones included⁴. A script that retries a failed upload ten times can spend up to ten upload calls on one video.
• Paging. Each extra page of results is charged again², so a sync job that walks a big channel's video list every few minutes adds up.
• Thumbnails and edits on top of uploads. One upload plus a thumbnail plus a metadata fix is 1 call from the uploads bucket and 100 units from the shared pool.
• Several apps on one project. A test harness and the production pipeline on the same project eat from the same 100 uploads.
• Multi-tenant tools. If you built something your clients use, all of their channels upload through your one project.

None of this is a bug in your code. The default is sized for light use, and real automation outgrows it.

Your three options

Uploading through PublishQ

Three routes reach the same API. Pick the one that matches how your pipeline already runs.

From the terminal or CI

npm i -g @publishq/cli
pq login
pq accounts list --platform YOUTUBE
pq media upload --file ./video.mp4
pq posts create --content "Weekly build log, episode 12" \
  --account <account-id> --media-id <media-id> --publish-now

Every command returns JSON, so a shell script or a CI job can read the IDs with jq. In CI, set PUBLISHQ_API_KEY=pq_live_... instead of logging in. Swap --publish-now for --scheduled-at 2026-12-01T09:00:00Z to queue it, or leave both off to keep a draft. The full command list is in the CLI docs and on the YouTube CLI page.

From your own code

Create a key at https://publishq.com/app/api-keys, upload the file to POST /media (or get a signed upload URL from POST /media/uploads for files up to 4 GB, see media uploads), then create the post:

curl -X POST https://publishq.com/api/v1/posts \
  -H "Authorization: Bearer pq_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "content": "Weekly build log, episode 12",
    "mediaIds": ["3b9f6c2e-71d4-4a8e-9c05-d2e8a6f41b77"],
    "accounts": [{
      "accountId": "a7c41e90-5d2b-4f63-8e1a-0c9b7d3f6e24",
      "platformSpecificSettings": {
        "description": "Changelog and repo links below.",
        "privacyStatus": "unlisted",
        "categoryId": "28",
        "madeForKids": false
      }
    }],
    "publishNow": true
  }'

content is the YouTube title (100 characters max). privacyStatus defaults to public, so set it if you want an unlisted test first. The rest of the YouTube fields, including thumbnailMediaId for a custom thumbnail, are on the YouTube API docs page. TypeScript users can use @publishq/sdk instead of raw HTTP. More on the REST side is on the YouTube API page.

From a coding agent

If you wrote your upload script with Claude Code, Cursor or Codex, you can hand the job back to the same agent. In Claude Code:

claude mcp add --transport http publishq https://publishq.com/mcp

Then run /mcp, choose Authenticate and approve in the browser. Cursor takes {"mcpServers":{"publishq":{"url":"https://publishq.com/mcp"}}} in ~/.cursor/mcp.json. For Codex, run npx @publishq/cli@latest login, then codex mcp add publishq -- npx -y @publishq/mcp@latest. After that, "upload ./out/ep12.mp4 to my YouTube channel as unlisted, category 28" is a whole request. The agent calls publishq_upload_media and publishq_create_post for you. There is a longer walkthrough in uploading to YouTube from Claude Code and on the YouTube MCP page.

The API, CLI and MCP server are on every plan, the Free plan included. See pricing for the rest.

Setup guides for other agents

Each of these pages has the setup for that agent and YouTube-specific prompts:

For any other agent, open https://publishq.com/agents and paste the line at the top of that page into your agent. It sets itself up and sends you a link to approve the login in your browser.

After the switch

The nightly job that used to stop at video 100 now finishes the queue. A retry no longer spends a call from a 100-a-day bucket of yours. The same create call can also carry TikTok, Instagram and Facebook accounts, each with its own caption, so one render can go out everywhere. If you want to keep a calendar view of what is queued, the YouTube scheduler shows it, and YouTube limits lists the title, description and tag rules a post still has to meet.

Stop planning uploads around Google's reset clock. Create a PublishQ account, connect your channel and send the next video through it.