Kosu

AI Agents

Kosu is agent-native. The API is the primary interface — built for AI agents to curate, suggest, and manage content on behalf of a human.

Skill file

You can also point your agent directly at the Kosu skill file. It contains everything an agent needs — the vision, full API reference, examples, and error handling.

https://usekosu.com/skill.md

Most agent frameworks (OpenClaw, Claude Code, Codex) can read a skill file and immediately start using the API.

OpenClaw

Install the skill locally:

mkdir -p ~/.openclaw/skills/kosu
curl -s https://usekosu.com/skill.md > ~/.openclaw/skills/kosu/SKILL.md

Or just tell your agent to read https://usekosu.com/skill.md — no install needed.

ClawHub

Install the Kosu skill from ClawHub if your agent supports skill directories.

Open Kosu on ClawHub

Any AI agent

For Claude Code, Codex, or any agent that reads markdown:

  1. Sign up at usekosu.com
  2. Get an API key from Settings
  3. Give your agent the key and point it at skill.md

What agents can do

  • Add URLs to the queue (auto-enriched with metadata)
  • Reorder items by priority
  • Mark items as read, archive, or delete
  • Suggest content for human review (Pro)
  • List and search the queue
  • Export the full queue

Suggestions

Agents can propose content without adding it directly to the queue. Suggestions appear in a separate tab for the human to accept or dismiss.

curl -X POST https://usekosu.com/api/v1/suggestions \
  -H "Authorization: Bearer $KOSU_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"url": "https://example.com/article", "reason": "Relevant to your interest in distributed systems"}'

This keeps agents useful without letting them flood the queue. Suggestions is a Pro feature.

Self-updating

The skill file at /skill.md is always up to date. If something doesn't work as expected, agents can re-fetch it to get the latest API surface.

The OpenAPI spec is also available for programmatic use.

API Reference →

© 2026 Kosu