Productclank Content Studio

ProductClank Content Studio

Let your agent set up and run your own content pipeline on ProductClank. It onboards the brand in a short interview (voice, audience, platforms, post types, the topics you want to own), then drafts posts in that voice, reads back how the reviewer scored them, nudges a draft with one word (shorter, punchier, deeper…), stages what you approve, and turns your reactions into standing rules so the next drafts are better.

The agent proposes; you dispose. Nothing is ever auto-published — you stage in the chat or the web tool, and publishing is your step there.

Not a community campaign. This is your own content pipeline. If instead you want the community to create content for a product (and reward them), use the Content Campaign capability in the productclank-campaigns skill.

When to use this skill

• "Set up content for Acme — ask me what you need to know about our voice and topics."
• "Draft three build-in-public posts about today's release into my content space."
• "Show me the queue. Approve the first one, make the second punchier, and remember: never say 'leverage'."
• "Turn this changelog into content candidates for TipRanks."

Prerequisites

1. A connected ProductClank account — via the Claude connector (OAuth) or an agent API key (Authorization: Bearer pck_live_…).
2. A content-enabled space — the agent can create one: POST /content/workspace turns the engine on for an Amplify space the user owns / delegates / manages, or creates a solo space for the brand with new_space:{name}. (The web alternative is https://app.productclank.com/content.)
3. Credits only for AI rewrites — setup, drafting, topics and reading the queue are free; revise / fix / humanize / review cost 2, a KB rule 1, a pasted brand doc 5.

Authentication

Every endpoint requires Authorization: Bearer <api_key>. Trusted agents acting on behalf of a user pass caller_user_id (the Claude connector sets this automatically from the OAuth session).

Identity — which account are you acting as?

This is the #1 thing to get right (and the most common source of "it returns nothing"). Every call runs as a ProductClank user, and you only ever see that user's spaces. There are two ways your agent maps to a user:

• Non-trusted agent (the default when you self-register). You act as your own linked user — so you only see spaces your agent's own account owns/manages, which for a fresh agent is none. To act for a human, link your agent to their account: POST /api/v1/agents/create-link → they open the returned URL and approve. After linking, every call acts as that user with no caller_user_id needed.
• Trusted agent (granted by ProductClank; e.g. a multi-user connector). You must pass caller_user_id — the ProductClank user id of the person you're acting for — on every call, and that user must have authorized your agent.

Building a connector for many users? Prefer the OAuth connector, which resolves the user automatically so you never handle caller_user_id at all.

The flow

1. Find the space — GET /content/spaces lists content-enabled spaces ({ space_id, name }). Empty? Offer to set one up (step 2). Otherwise GET /content/workspace?space_id= to read the brand's calibration.
2. Onboard (first time only) — interview the user ONE question at a time: brand & site → audience → tone (offer the archetypes: Professional Authority · Friendly Educator · Conversational Peer · Inspirational Motivator · Analytical Thinker · Storyteller) → 2–3 example posts → platforms → post types → 3–8 topics. Read it back, then call POST /content/workspace once with the answers as fields (free). new_space:{name} creates a solo space when the user has none.
3. Draft in the voice — POST /content/candidates with 1–25 candidates written IN the workspace's voice, on its topics, each tagged with one of its platforms. They are auto-scored by the reviewer a moment later.
4. Present the queue — GET /content/queue?space_id= returns each draft with review.score, verdict, and the reviewer's one-line summary. Show them one by one: approve, tweak, or drop?
5. Approve / tweak — PATCH /content/drafts: stage (approved), discard, revise with a preset (shorter, longer, punchier, deeper, simpler, more_specific, less_salesy, more_casual, more_formal) or the user's words, fix (reviewer's notes), humanize, review, edit.
6. Teach — when a reaction should apply to every future draft, POST /content/feedback turns it into a rule in the brand voice KB (applied immediately by default). Say in one line what was learned.
7. Human publishes — staged drafts wait in the web tool's All Content queue (review_url). Publishing is the user's step there.

Full request/response shapes for every endpoint: references/API_REFERENCE.md.

1. List content spaces

GET /api/v1/agents/content/spaces

{
  "success": true,
  "spaces": [
    { "space_id": "c75562db-…", "name": "ProductClank Community" },
    { "space_id": "49398afc-…", "name": "TipRanks" }
  ]
}

Empty list → the user hasn't enabled the content engine on any space yet. Point them to https://app.productclank.com/content.

2. Write content candidates

POST /api/v1/agents/content/candidates

{
  "space_id": "c75562db-3341-…",
  "candidates": [
    {
      "text": "Shipped agent-written content today 🚀 Your agent can now draft posts straight into your ProductClank pipeline for review.",
      "title": "Agent content candidates launch",
      "platform": "ProductClank X",
      "template": "Proof Point"
    },
    { "text": "A human still reviews and schedules everything — nothing auto-publishes." }
  ]
}

Only text is required per candidate. platform must be one of the workspace's platforms (omit it to use the first; an unknown label is rejected with 400 unknown_platform rather than mis-tagged). template defaults to Build-in-Public. Drafts are auto-scored against the brand voice in the background — read the scores via GET /content/queue.

Response:

{
  "success": true,
  "created": 2,
  "draft_ids": ["9416624f-…", "9ad39a19-…"],
  "space_id": "c75562db-3341-…",
  "review_url": "https://app.productclank.com/content?space=c75562db-3341-…"
}

Share review_url with the user so they can review, edit, and schedule the drafts.

Fields

Limits & guardrails

• Up to 25 candidates per call. Split larger batches across calls.
• Free to draft — no credit charge; drafts are auto-scored (score-only, never auto-staged).
• Never auto-published. Candidates stop at the queue; the user stages, and publishes from the web tool.
• Stage and discard are the user's decisions — act on their say-so, never on your own judgment.
• Writes are gated to spaces the user actually controls, and honor authorization: if the user revokes your app in Connected Apps, writes stop immediately.

Errors

Troubleshooting

Most "it returns nothing / it says I need access" problems are an identity mismatch (see Identity), not a missing approval.

⚠️ Do NOT call POST /api/v1/agents/authorize to fix "I can't see the spaces." That endpoint is a trusted-agent self-grant that takes a user_id (not an agent_id), and it does not substitute for a missing caller_user_id or for linking. Reach for create-link (to act as a user) or caller_user_id (if trusted) instead.

Safety

The text you write is a draft — review the content you generate before writing it, and confirm the target space with the user. The user is always the last step: they review, edit, and schedule everything from the content tool. The platform never auto-posts.