Upload from scripts and CI

The REST API takes the same API key as MCP and acts on the same namespace. Base URL: https://app.jitprompt.com.

Upload

curl -H "Authorization: Bearer $JITPROMPT_KEY" \
  -F "file=@docs/refund-policy.md" -F "title=Refund policy" \
  https://app.jitprompt.com/documents
# {"document_id": "…", "status": "pending", "deduped": false}

Uploading identical content again returns the existing document with "deduped": true and uses no ingestion — so re-running the upload of an unchanged file is free.

Update in place

Replacing content keeps the document’s id and title:

curl -X PUT -H "Authorization: Bearer $JITPROMPT_KEY" \
  -H "Content-Type: text/markdown" --data-binary @docs/refund-policy.md \
  https://app.jitprompt.com/documents/$DOC_ID/content

Wait for it

curl -s -H "Authorization: Bearer $JITPROMPT_KEY" \
  https://app.jitprompt.com/documents/$DOC_ID | jq -r .status   # … ready

GitHub Actions

name: publish instructions
on:
  push:
    branches: [main]
    paths: ["instructions/**"]
jobs:
  publish:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - name: Upload changed files
        env:
          JITPROMPT_KEY: ${{ secrets.JITPROMPT_KEY }}
        run: |
          for f in instructions/*.md; do
            curl -sf -H "Authorization: Bearer $JITPROMPT_KEY" \
              -F "file=@$f" -F "title=$(head -1 "$f" | sed 's/^# //')" \
              https://app.jitprompt.com/documents
          done

New content creates a new document; to keep one document per file, store the document_id and use PUT …/content instead. Every upload or replacement counts as one ingestion against your plan.

The full endpoint list is in the API reference.