Piper Morgan Beta Testing Guide

Version: 0.9.0.0 Last Updated: October 31, 2026 For: Private Beta Testers

Rewritten 2026-09-21 (Lead), closing #1804. This guide previously walked testers through a local clone/Docker/setup-wizard install — a flow retired by #1708 (2026-08-31): Piper Morgan’s beta is a hosted app at beta.pipermorgan.ai, and that’s what this guide now describes. Engineers who want to run the code locally: see CONTRIBUTING.md — local setup lives there, not here.


Returning Tester? Start Here

If you already have an account, skip straight to what matters:


Quick Navigation

Section Description Start Here If…
Chapter 1: Setup Account creation, your LLM key, optional connections You’re setting up for the first time
Chapter 2: Testing Test scenarios, features to explore You have an account and want to start testing
Chapter 3: Troubleshooting Common issues and solutions Something isn’t working

Chapter 1: Setup

Piper Morgan runs as a hosted app — there is nothing to install. Setup is three steps: an account, your LLM key, and (optionally) your tool connections.

What you need

Optional (for the integration features):

Time commitment: ~10 minutes to first conversation; 15–30 minutes weekly for feedback during the beta period.

Step 1 — Create your account

  1. Go to beta.pipermorgan.ai and choose “Create your account”
  2. Choose a username and password, and enter your email
  3. Enter your invite code — account creation is invite-gated; the form won’t proceed without a valid, unused code
  4. Log in at beta.pipermorgan.ai

Step 2 — Add your LLM key (the one required step)

Piper can’t answer anything substantive until it has a key of yours to run on:

  1. Open Settings → LLM API Keys
  2. Paste your Anthropic key and save

Until you do this, Piper will tell you so honestly — a keyless “hi” gets a greeting and a pointer to Settings, and a keyless request gets a refusal that names exactly what’s missing. That’s by design (see What’s New): no key, no spend, no borrowing.

Step 3 (optional) — Connect your tools

In Settings → Integrations: GitHub and Google Calendar. Each is optional; core chat, lists, files, and standups work without either. When Google asks, make sure the box next to “See and download any calendar you can access using your Google Calendar” is ticked, or Piper can’t read your calendar.

Important Disclaimers — Please Read

⚠️ PRIVATE BETA SOFTWARE WARNING ⚠️

This is private beta software (version 0.9.0.0). By proceeding, you acknowledge:

  1. Expected Issues: bugs, rough edges, and incomplete features are normal
  2. Data Loss Risk: your data on the beta instance may be lost at any time
  3. No Production Use: do NOT use it for mission-critical or time-sensitive work
  4. API Charges: you are responsible for all LLM API costs your key incurs
  5. Security: not security audited — use test data only, no sensitive information
  6. No Warranty: provided “as-is”, without any warranty whatsoever
  7. No Support SLA: best-effort support only, no guaranteed response times

See the Terms at https://pipermorgan.ai/terms/ for complete terms.


What’s New in 0.9.0

v0.9.0.0 is the first private beta release and the one that closes the MVP milestone. It was cut on 2026-10-31, after v0.8.14.0 (2026-09-23). Theme: trust in what Piper tells you. Most of the 361 issues closed since 0.8.14 are cases where Piper said something that was not true or did not say something that was. Each now has a fix, and many have a regression test pinned to the exact wording that failed.

Setup works for an invited tester (#2078, #1875, #2089, #2176):

Integrations say what is true (#2012, #2029, #2045, #2071, #2148):

Piper tells you when it could not read something (#2182, #1965, #2072, #2076):

Todos, reminders, projects and GitHub issues in chat:

Feedback, privacy and your data:

In this beta: the connectors offered are GitHub and Google Calendar only. A hosted MCP server at mcp.pipermorgan.ai ships too, as private-beta work that is rougher than the web app.

See Release Notes v0.9.0.0 for full details, and releases/README.md for prior release history.


Chapter 2: Testing

This chapter covers what to test and how. If you’re already set up, start here.

What to Test in 0.9.0

Trust is the focus. Does setup finish for you? Does every status you see match what is really happening? When Piper cannot read something, does it say so? Send the exact wording of anything that looks wrong. Each report becomes a test row.

The six walks below are suggested scenarios built from this release’s fixes. They have not yet been walked by the team, so treat a surprise as a finding, not as your mistake.

Test Walk 1: Setup as an invited tester (#2078, #1875)

  1. Create your account at /setup with your invite code and run the setup wizard
  2. Check: the wizard finishes without “Setup Failed”, and your key is stored
  3. Check: you do not see the local System Requirements check or Slack and Google Calendar app-credential forms
  4. Start a conversation before adding a key, then add one. Check the conversation is still in your sidebar

Test Walk 2: Integrations say what is true (#2012, #2045, #2148)

  1. Under Settings → Integrations, connect GitHub and Google Calendar and run Test on each
  2. Check: the card updates its status and “last checked” line, and the “healthy” count follows
  3. If you can, revoke the GitHub grant on GitHub and run Test again. Check it reports the connection as broken, not successful
  4. Start the Google Calendar connect and leave Google’s consent screen without finishing. Check Piper does not say “Calendar Connected”
  5. Ask “what’s my agenda Thursday?” and check it honors the weekday

Test Walk 3: Failed reads are reported (#2182, #2076, #2173)

  1. Ask for a standup or “what’s on my radar” while GitHub or your calendar cannot be read (for example with a revoked grant)
  2. Check: Piper says it could not read the source, rather than showing a free day or “no activity”
  3. With no GitHub username set, check the reply does not list every open issue as your own work
  4. Ask “what changed since yesterday?” and check the reply explains the window it read

Test Walk 4: Todos and reminders (#2094, #1987, #1985)

  1. Say “Add three todos for the launch: draft brief, book venue, send invites”
  2. Save a reminder, then say “delete that reminder”. Check it removes the one you just saved
  3. With some overdue todos, ask for your todos. Check Piper never says the list is clear
  4. ⚠️ Known: “Delete todo 1 and 2” removes only the first (#2025). Report anything beyond what the Known Issues page lists

Test Walk 5: GitHub issues in chat (#2205, #2115, #1941)

  1. As a first action, say “Create an issue” and check Piper asks about the repository rather than choosing one for you
  2. Right after creating one, say “Show me that issue” and check it shows the issue
  3. Say “close issue 99999 in /" (a number that does not exist). Check Piper says there is no such issue and nothing was changed

Test Walk 6: Your data (#2065, #2096)

  1. Hide or delete a chat and check it leaves your sidebar
  2. Use Send feedback from inside the app, including from the setup wizard or an error page
  3. Check Settings does not promise Export or delete actions the page lacks

Basic Functionality Tests

  1. Basic Chat: “Hello, what can you help me with?”
  2. Task Creation: “Add a todo: Review Q3 metrics”
  3. Information Query: “What tasks do I have?”
  4. File Upload: upload a PDF or DOCX (max 10MB) and ask for analysis
  5. Document Summary: “Summarize the document I just uploaded”
  6. Multi-User Privacy: if testing alongside someone else, verify you can’t see each other’s data

Exploring Piper’s Features

All URLs below are on the hosted app — e.g. https://beta.pipermorgan.ai/lists.

Lists, Todos, and Projects

  1. Create a List: go to /lists → “Create New List” → name it “Alpha Testing Tasks” → verify it appears
  2. Edit and Delete: open your list, edit the name, then try deleting one (should confirm first)
  3. Repeat for Todos and Projects (/todos, /projects) — the three resource types should behave consistently

File Management

  1. Upload: /files → upload or drag-and-drop (PDF, DOCX, TXT, MD, JSON; max 10MB)
  2. Download and Delete: both from the file’s row
  3. Privacy: files are private to you — another tester should never see them

Interactive Standup Assistant

Worth trying alongside the walks above (suggestion chips such as “nothing” and “skip” now appear, #2027):

Authentication

  1. Logout: user menu (top right) → Logout → verify redirect to login
  2. Login again: your data (lists, files) should all still be there

Sharing & Permissions (advanced — needs two accounts)

  1. Share a list: open it → “Share” → another tester’s email → role “Viewer” or “Editor”
  2. Verify the viewer can see but not edit; the editor can edit
  3. Conversational forms work too: “share my Alpha Testing Tasks list with [email] as editor”, “who can access my Alpha Testing Tasks?”

Chapter 3: Troubleshooting

Login & account issues

Invite code rejected? Codes are single-use and exact — check for copy/paste whitespace. If it still fails, yours may already have been consumed; reply to your invitation email and ask for a new one.

Forgot your password? Reply to your invitation email. There’s no self-serve reset in the beta yet.

Can’t reach the site? Check https://beta.pipermorgan.ai/health — if it doesn’t return a healthy JSON response, the instance is down; please report it.

Key & chat issues

“Piper runs on an LLM key of your own…” on every message? You haven’t added a key yet (or it didn’t save) — Settings → LLM API Keys. This message is working as intended for keyless accounts.

Honest key error after adding a key? Verify the key works in your provider’s own console; regenerate it if in doubt. Anthropic keys start sk-ant-.

“Something unexpected happened”? That generic message on a key-shaped problem is itself a bug we’re hunting (#1824) — please report the exact message and what you did.

High API costs? Piper runs on your own Anthropic key. Watch your provider dashboard; concise-preference settings reduce token use.

Feature issues

Standup seems stuck? Say “start over” (genuinely restarts now) or “not now” (lets go of the conversation). If a standup ever shows content you didn’t say, that’s a serious bug — please report the transcript.

File upload fails? Formats: PDF, DOCX, TXT, MD, JSON; max 10MB; you must be logged in.

Sharing fails? The other account must already exist, and you must own the resource.


Providing Feedback

We need your feedback to improve! Please report:

What to Report

How to Report

  1. Private: email support@pipermorgan.ai. The team reads it, including xian, and it stays with us unless you ask us to share it.
  2. Public: a GitHub issue, for anything you are happy to have public. A GitHub issue is public, so leave out anything you would not post on a public page.
  3. Weekly check-in: optional 15-minute calls available

Helpful Feedback Format

WHAT I TRIED: [specific action, ideally the exact message you sent]
WHAT I EXPECTED: [expected result]
WHAT HAPPENED: [actual result / exact reply text]
ERROR MESSAGE: [if any]
SEVERITY: [blocker/major/minor]

Privacy & Data Collection


Questions?

This is private beta software (version 0.9.0.0) — expect rough edges, and thank you for being an early adopter. 🚀


See Also


Last updated: October 31, 2026 Software version: 0.9.0.0