ClaprDocs
Sign in

Clapr for agents

Clapr is built agent-first: everything a human can do on the board, an agent can do through the MCP server or the REST API. This page gives your agents the conventions to do it well — draft boldly, and leave the shipping decisions to humans.

The working agreement

The board is the review interface between your agents and your team. The conventions that make that work:

  • Agents draft — ideas, titles, hooks, scripts go in as cards in the first column.
  • Humans review — agent-created cards carry an Agent badge; the creator edits and moves them forward.
  • Agents never publish — moving a card past editing is a human decision.
  • Data over vibes — agents check performance of published videos before proposing what's next.

AGENTS.md / CLAUDE.md snippet

Paste this into your repo's AGENTS.md (or CLAUDE.md) after connecting the MCP server:

## Clapr — YouTube content board

This project uses Clapr (https://www.clapr.io) as the content command center.
You are connected via the "clapr" MCP server (setup: https://www.clapr.io/docs/mcp).

Conventions:
- Draft new video ideas as cards in the first column ("Idea") using create_card.
  Always fill title AND a short description with the hook angle.
- Never move a card past "Editing" — publishing decisions belong to humans.
- Before drafting new ideas, call get_performance on recent published cards and
  favor topics that performed well.
- When you change anything, add a comment (add_comment) summarizing what you did
  and what needs human review.

Claude skill

For Claude Code, you can go one step further and install this as a skill. Create .claude/skills/clapr-content-planner/SKILL.md in your project (or ~/.claude/skills/… for all projects) with:

---
name: clapr-content-planner
description: Draft, organize and review YouTube video ideas on the user's Clapr
  board via the clapr MCP server. Use when the user asks to plan videos, add
  ideas, check content status, or review video performance.
---

# Clapr content planner

You have access to the user's Clapr board through the "clapr" MCP server
(tools: list_workspaces, list_statuses, list_cards, get_card, get_performance,
create_card, update_card, move_card, add_reference, add_comment).

## Workflow

1. **Orient first.** Call list_workspaces and list_statuses before writing
   anything. Never assume column names.
2. **Draft into the first column.** New ideas go in as cards with a strong
   title, a one-line hook in the description, and a scheduled_date if the user
   gave one.
3. **Ground ideas in data.** When suggesting topics, check get_performance on
   recent published cards and say why each idea should work.
4. **Humans approve.** Never move cards past the editing stage. Your job ends
   at a well-prepared draft; the creator and editor review and ship it.
5. **Leave a trail.** After a batch of changes, add_comment on the affected
   cards (or report in chat) what you did and what needs review.

llms.txt

Agents browsing the web can read https://www.clapr.io/llms.txt for a compact, machine-readable overview of Clapr and its interfaces (/llms-full.txt for the long version).