For AI agents

A concise, accurate setup + usage reference for an AI client connecting to Drp. A plain-text copy is at /llms.txt.

Connect

MCP server (Streamable HTTP): https://mcp.drp.dev

• OAuth (preferred): add the URL with no key; on a 401 the server returns WWW-Authenticate pointing at https://api.drp.dev (authorization server, dynamic client registration + PKCE). Complete the browser login; use the access token as the bearer.
• API key: send Authorization: Bearer drp_live_… (created by the user at drp.dev/app/settings/keys).

First, orient yourself

• Call whoami → confirms the account and the default team drops land in.
• Call list_teams → team slugs you can post to. list_folders → folders.

Post content (base64 is the normal path - Drp hosts it; do NOT host elsewhere)

• post_image { imageBase64, contentType, title?, visibility?, team?, folder? }
• post_markdown { body, title?, … }
• post_html { body /* full HTML, runs sandboxed */, … }
• post_code { filename, content /* text, highlighted read-only */, … }
• upload_file { filename, contentBase64, contentType, … }
• post_report { body /* with  */, images:[{slot,imageBase64,contentType}], … }

Iterate on a document (research-as-you-go)

• get_drop { dropId } → read back the current markdown body + filled image slots.
• append_drop { dropId, content } → add markdown to the bottom using a drop ID or Drp link, without a full rewrite.
• update_drop { dropId, body?, title?, visibility?, folder? } → replace the full markdown body or edit metadata.
• add_report_image { dropId, slot, imageBase64, contentType } → add an image. If the body has  it fills that spot; otherwise it's appended to the bottom (markdown auto-upgrades to a report).

Pattern for running notes and completed-item lists: post_markdown once, keep the dropId, then call append_drop for each addition. Use update_drop only when the full body needs to change.

Rules that matter

• Destination: omit team = default/personal; pass a team slug to target a team. Default visibility is team; use public for shareable links / changelog.
• Folders auto-create: just pass folder: "Name".
• Size: ~24 MB inline soft cap, 100 MB hard. For many/large report images, create the report then call add_report_image once per image. Don't go find a URL - send the bytes.
• Result: every post returns a drp.dev/r/<slug> URL - surface it to the user.
• Reports with a missing image return incompleteSlots + a hint; follow it.

Review, evidence, and release workflows

• Drp Review: call review_changes with the real unified diff. It returns validated findings, merge risk, and an evidence plan without storing anything by default.
• Drp Evidence: capture and judge the mandatory items from that plan, then use post_report for the private internal report.
• Drp Release: use duplicate_drop to preserve the internal source, rewrite the copy for users, run redact_content, and make only the copy public.
• Drp Full: ask your agent to “use drp-full”. It can discover drp_full and follow the complete review-to-release execution contract.

Example: a PR-delivery report

post_report({
  title: "PR #42 - streaming uploads",
  visibility: "public",
  folder: "PRs",
  body: "# What changed\n![dashboard](drp:dash)\n## Before / after\n![before](drp:before) ![after](drp:after)",
  images: [
    { slot: "dash",   imageBase64: "…", contentType: "image/png" },
    { slot: "before", imageBase64: "…", contentType: "image/png" },
    { slot: "after",  imageBase64: "…", contentType: "image/png" }
  ]
})