Quick Start

Two steps for everyone: connect your AI tool, then say aswritten. Your AI loads the perspective, notices when the protocol isn’t installed where you are, and tells you what it needs. Invited users are done in minutes; building from scratch continues into your first memory.

What You Need

An AI tool with MCP support: Claude Code, Claude Desktop, GitHub Copilot, Cursor, n8n, or any MCP-compatible client
Invited users need nothing more; “build your own” needs a GitHub account (or we’ll manage the repo for you on Expert)

Step 1: Connect your AI tool

Same for every path.

Claude Code

Run this once from anywhere — it registers aswritten across all your projects:

claude mcp add --scope user --transport http aswritten https://api.aswritten.ai/mcp

Open Claude Code in any project directory and run /mcp. Select aswritten to authenticate (opens a browser). Restart Claude Code so the connection takes effect.

Claude Desktop

Settings → Connectors → Add custom connector. Name it aswritten. Server URL: https://api.aswritten.ai/mcp. Press Connect and authenticate in the browser.

Claude Desktop needs two things. Both matter — neither alone is enough:

A Project whose custom instructions carry the aswritten protocol. Type aswritten in a chat and your AI hands you the file to paste — project-instructions.txt — whole, into the Project’s custom instructions. You don’t need repository access; the file is the same for every perspective. Without it, Claude sees the tools but not the way of working with them.

Start every chat from inside the Project (sidebar → the project → New Chat). Chats started from the home screen reach the tools but skip the protocol.

Other MCP-compatible tools (GitHub Copilot, Cursor, n8n)

Add an MCP server with URL https://api.aswritten.ai/mcp. Authenticate with your email. Restart your AI tool so the connection takes effect.

Microsoft 365 Copilot is a different product and is not MCP-compatible. If that’s all you have, install Claude Desktop.

Login vs. sign-up. When the browser window opens to authenticate:

If you were invited to a team — log in with the email from your invite. You’re already on your team’s tier; there’s nothing to pick, and no card to enter.

If you’re starting fresh — start an Expert trial to build your own (a card is required up front), or Team to roll out to others. The trial is the real product, not a tour: it lets you add your own knowledge, which is the entire point. See Pricing.

Step 2: Say `aswritten`

Open a session and type aswritten. Nothing is written until you say yes.

Claude Code. Open the folder you work in — your perspective’s repository, or any folder — and type aswritten.

In your perspective’s repository, your AI installs the protocol into the repository’s CLAUDE.md.
In any other folder, it asks which you mean: make this folder a new perspective, or connect it to an existing one — and which repository. If you know the repository, say it in one line: owner/repo aswritten. Your AI writes the folder’s CLAUDE.md with the repository named above the protocol block, and sessions in that folder load that perspective from then on. Branches are main unless you name one.

Claude Desktop. Create a Project, start a chat inside it, type aswritten. Your AI tells you the protocol isn’t installed and hands you project-instructions.txt to paste into the Project’s custom instructions. Start chats inside the Project.

Step 3: Pick your path

Path A: I was invited to a team perspective

Your team admin has set up the perspective — nothing to scaffold, no first-memory interview, no GitHub App to install. Your admin gives you the repository name (owner/repo).

Claude Code: in any folder, owner/repo aswritten. Or clone the repository and start Claude Code inside it — the protocol is already in its CLAUDE.md.
Claude Desktop: the Project and the paste, as in Step 2.

Then ask away. Type aswritten to load the perspective and ask whatever you’d ask a senior person on the team. Citations show where each answer came from. For question patterns that get good answers, see Asking the Perspective Well.

No GitHub access? That’s the common case. You don’t need GitHub to use the perspective — being added to the team is what gives you load/ask/remember access. GitHub only matters for merging the PRs that result from saving memories (the team admin usually does this) and for the Claude Code workflow that clones the repo locally.

Path B: I’m building my own perspective from scratch

Start an Expert or Team trial in the auth window from Step 1. Then in your AI tool, type aswritten. Claude walks you through:

Connect GitHub — install the aswritten GitHub App and select which repos to connect (Expert: we manage the repo for you, no GitHub needed; Team: connect your own GitHub org)
Initialize — scaffold the aswritten files (ASWRITTEN.md, workflows) via a commit to your default branch
First memories — from your call transcripts if you have them (fastest), or an interview (next sections)

If you’re using Claude Desktop, also paste project-instructions.txt into a Project’s custom instructions (Step 2).

The fast path: start from your call transcripts

If you record your calls — Fireflies, Zoom, Google Meet, Gemini notes — you’re already sitting on the source material. You don’t have to write anything.

Paste a transcript into the chat and ask your AI to draft a memory from it. It distills what the call actually holds — the decisions, positions, and reasoning, with who-said-what and the exact quotes preserved.
Add your commentary before it saves. “What here is settled? What’s already out of date? What did you decide after this call?” Your annotation is what turns a record into a perspective.
Repeat for your last handful of calls. Each one goes through the same draft → review → save → PR cycle. A working first perspective from five or ten transcripts is an afternoon, not a quarter.

Have months of recordings? Don’t go one-by-one. Ask your AI to look across the whole set, surface the recurring topics and pivotal moments, and draft a small collection of memories — one per theme, each carrying the key quotes from across the calls, every quote attributed to who said it, when, and which call it came from. You review the collection once, not every transcript — and citations still land on the original conversation.

Your AI working sessions are source material the same way. At the end of a Claude Code session where something real got decided, ask it to save a memory of what was decided and why — the session itself is the transcript.

Your first memory (Path B)

No transcripts handy? Claude interviews you instead — about your project, key decisions you’ve made, who’s involved, what’s in progress. Talk naturally; Claude drafts a memory document and asks you to approve it before saving.

What to talk about:

What does your project/org do? Who’s involved?
The 2-3 most important decisions you’ve made recently — and why, not just what
What do new people always ask about? What takes the longest to explain?
The stuff that lives in your head and never got written down

What happens after you approve:

Extraction — runs during the save (~2-3 minutes). An LLM reads your memory and extracts structured knowledge: decisions, actors, relationships, how settled each claim is.
PR opens — showing exactly what was extracted. Every fact traces back to your words. Review and adjust if needed.
Merge — then reload the perspective. Ask a question about your project: if the answer cites your decisions with provenance instead of giving generic advice, it’s working.

From here the pattern repeats: decisions come up in conversation → Claude offers to save them → extraction → PR → merge → the perspective grows. Use introspect in interview mode to see which domains are thin and get targeted questions to fill them.

Coming from a sales call? Your first memory was already created during that conversation. Load the perspective, see what’s there, and build from it.

Adding teammates

On a Team or Organization plan:

Add each teammate to your team (manage-account) — this is what grants them access
Each teammate connects the MCP server (Step 1) and logs in with their invited email
They say aswritten with the repository name (Step 2, Path A) — on first load they have everything the team has documented

Teammates can save their own memories. Every addition arrives as a PR, so knowledge is reviewable before it enters the perspective. Different people contribute different domains — your architect documents architecture, your PM documents strategy.

Multiple repos

aswritten works across repos and organizations. Use switch-repo to change which repo Claude is working with. Each repo stores its own perspective.

Troubleshooting

“aswritten isn’t responding / answers feel generic” — On Claude Desktop this is almost always the Project gotcha: project-instructions.txt not pasted, or the chat started from the home screen instead of inside the Project. See the Step 1 Desktop callout. In Claude Code, a folder without the protocol tells you so when you type aswritten — say yes to the install. Otherwise check authentication (/mcp in Claude Code, Settings → Connectors in Desktop).

“The perspective is empty” — You haven’t saved any memories yet, or the PR hasn’t merged. Save your first memory and merge the PR.

“The perspective is sparse” — Check for pending shares to import (the AI checks automatically, or run import with no arguments). If not, use introspect in interview mode to identify what domains need memories.

“Extraction is taking too long” — It runs inside the remember call and typically takes ~2-3 minutes. If it seems stuck, check the tool output for errors.

“I merged the PR but the perspective hasn’t changed” — The AI caches the perspective. Ask it to reload explicitly, or start a new session.