Open And Async Server

Bring async-first working practices into your editor and AI assistant. This Model Context Protocol server gives your AI the tools to draft a decision doc, turn a meeting into an async artifact, pressure-test a status update, or settle a sync-vs-async debate — without leaving the tools you already work in.

It's the method from Open and Async — the collaborative software-development playbook for remote and distributed teams — as working tools, not reading.

What you can do with it

• Stop writing decision docs from a blank page. Hand your AI the decision and the options; get back a structured ADR scaffold — context, tradeoffs, the call, and how reversible it is.
• Turn "let's hop on a call" into an artifact. Give it a meeting's purpose and agenda; get the async equivalent with an owner and a deadline.
• Ship status updates that don't blindside anyone. Score a draft against the "work loudly / no surprises" rubric and get concrete fixes before you post it.
• Settle sync vs. async in seconds. Describe the task; get a recommendation and the rule behind it.
• Run a standup without a meeting. Drop in a structured async-standup template your team can adopt today.
• Win the "async is too slow" argument. Map common objections to a ready reframe, pull role-aware guidance, or search the book's principles — each answer cited and linked.
• Get coached end to end. Run the coach prompt with a situation and it triages sync vs. async, generates the right artifact, adds role guidance, and caps with a shareable line — one composed answer, or a nudge to just have the conversation when that's the better call.

Install

Claude Code:

claude mcp add open-async -- npx -y @open-and-async/mcp

Claude Desktop (claude_desktop_config.json):

{
  "mcpServers": {
    "open-async": {
      "command": "npx",
      "args": ["-y", "@open-and-async/mcp"]
    }
  }
}

Any MCP client that speaks stdio works the same way: run npx @open-and-async/mcp.

The toolbox

Method tools — pure utility, use them on any project

Reference tools — the book's thinking, on demand

Resources

• book://outline — sections, chapters, TL;DRs (JSON)
• book://taglines — taglines + quote-card URLs (JSON)
• book://about — what this server is, what it does and doesn't contain, and where to get the book

Prompts

Invoke these directly from your client (e.g. as slash commands):

• coach — the full method on one situation. Triages sync vs. async, generates the right artifact, enriches with role guidance, and caps with a shareable tagline — one composed deliverable. Stops and recommends a real-time conversation when the situation genuinely needs one.
• async-standup, write-adr, meeting-to-issue, weekly-update — parameterized templates for a single task.

Good to know

It's a real tool, not a paywall. The method tools work on their own — no book required. The reference tools answer from the book's already-public summaries and a paraphrased framework layer, always capped and always cited so you can trace any snippet to its chapter. The "get the book" link surfaces about once per session, not on every line — it's a tool, not an ad. No verbatim book prose is bundled (the only data file is data/book.json.br), so what you install is genuinely useful, not a teaser.

Staying current. data/book.json.br ships with a version that tracks the book's edition, so the server can tell you which edition it's based on and re-sync when a new one lands.

Generated output is a template, not the author speaking. The method tools format your input into decision docs, standups, and triage calls. That output is generated guidance to adapt — it is not a personal statement, quote, or endorsement by Ben Balter or Open & Async LLC, and shouldn't be presented as one. Reference-tool snippets are summaries the tool cites back to the book, not verbatim prose.

Licensing

This package is split-licensed — see LICENSE for the overview:

• Code (everything under src/) — MIT.
• Data (data/book.json.br) — proprietary; © Open & Async LLC. You may use it only as part of running this software. No redistribution as a standalone dataset, no derivative datasets, no model training. See DATA-LICENSE.md.

The full work — the stories, the voice, the complete argument — lives in the book: open-and-async.com.