Markdown Skiller
Markdown skills, served to your AI agents over MCP

A skill library for your team.

Curate organizational know-how as markdown files. Control who reads what, comment inline, resolve threads, and expose it all to your AI agents through a Model Context Protocol server.

Sign in to your orgSee what's inside

Markdown-native · SSR-rendered · Built on Next.js 16 + Postgres

Everything you need to operationalize team knowledge

Markdown is the storage format. People review, refine, and own it. Your agents read it the moment it lands.

  • MCP server, ready out of the box

    Expose your skill library to Claude, ChatGPT, Cursor, or any MCP-aware agent. Tool calls return clean markdown so models can reason over your team's institutional knowledge in real time.

  • Robust user access controls

    Grant access at the folder or file level. Per-skill grants override folder-level access, so sensitive playbooks never leak — and every viewer is listed transparently on the page.

  • Tiered permission system

    Admins manage members and ownership; members read, write, and refine. Roles compose with per-folder grants, giving you flexibility without YAML soup.

  • Inline commenting and resolution

    Highlight any span of markdown, leave a comment, resolve when it's handled. Live highlights stay in sync between the editor and the rendered preview.

  • Live, debounced preview

    Edit on the left, watch the preview update on the right with a tight 100ms debounce. Hit Focus to dismiss the preview when you want a clean writing surface.

  • Folders the way you'd expect

    Self-referencing folder tree per organization, with breadcrumbs, drag-friendly URLs, and an explorer view that scales from a handful of skills to thousands.

From scratch pad to source of truth

A clear loop from authoring to agent-readable output.

  1. 01

    Write skills as markdown

    Capture how your team runs payroll, ships releases, or handles support — anything you'd otherwise lose to Slack history.

  2. 02

    Review and refine together

    Comment on tricky passages, resolve threads, and let access controls keep sensitive material in the right hands.

  3. 03

    Ship it to your agents

    Plug the MCP server into Claude or Cursor and let your agents read the same source of truth your humans already trust.

Model Context Protocol

Your agents already know how your team works.

Two tool calls — that's it. get_all_skills lists every skill the caller can see, organized by folder. get_skill returns one skill's full markdown. The MCP server respects your access controls, so agents see exactly what their human counterpart would.

  • Honors per-folder and per-file access grants
  • Returns clean markdown — no HTML cleanup needed
  • Scoped to a single organization per session
claude · mcp · markdown-skiller
> Tools available
  • get_all_skills()
  • get_skill(id)

> get_all_skills()
{
  "folders": [
    {
      "name": "Engineering",
      "skills": [
        { "id": "skl_82h", "name": "Cutting a release" },
        { "id": "skl_91k", "name": "Triaging on-call pages" }
      ]
    },
    {
      "name": "Sales",
      "skills": [
        { "id": "skl_55a", "name": "Discovery call playbook" }
      ]
    }
  ]
}

> get_skill("skl_82h")
# Cutting a release
1. Open the release dashboard…

Give your team — and your agents — one source of truth.

Stop pasting Slack threads into prompts. Stand up a skill library your humans maintain and your agents read.

Sign in to get started