Piper Morgan Alpha Testing Guide

Version: 0.8.14.0 Last Updated: September 23, 2026 For: Alpha 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 alpha is a hosted app at alpha.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:

• What’s New in 0.8.14 — your preferences persist, every clock face is on your zone and labeled, the agenda shows real times, one-line add-project, honest GitHub-write outcomes
• What to Test in 0.8.14 — priority test walks for this release
• Troubleshooting — if something isn’t working

Quick Navigation

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

• A modern browser
• Your invite code — alpha is invite-only; codes come directly from xian@pipermorgan.ai. If you don’t have one, that’s the address to write.
• At least one LLM API key of your own:
  • An Anthropic API key (Claude), and/or
  • An OpenAI API key (GPT-4-class models)
• Budget roughly $5–20 for LLM API costs over the alpha period — Piper runs on your key and your account; it never bills anyone else’s (see What’s New).

Optional (for the integration features):

• A GitHub personal access token (issue creation/management features)
• A Notion API key (document features)
• A Slack workspace you can connect (notifications, chat-with-Piper-from-Slack)

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

Step 1 — Create your account

1. Go to alpha.pipermorgan.ai/setup
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 alpha.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 Keys
2. Paste your Anthropic and/or OpenAI key and save
3. If you add both, you can pick a default provider; either alone works

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 (personal access token), Notion (API key), Slack (OAuth). Each is optional; core chat, lists, files, and standups work without any.

Important Disclaimers — Please Read

⚠️ ALPHA SOFTWARE WARNING ⚠️

This is pre-release alpha software (version 0.8.14.0). By proceeding, you acknowledge:

1. Expected Issues: bugs, rough edges, and incomplete features are normal
2. Data Loss Risk: your data on the alpha 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 ALPHA_AGREEMENT_v2.md for complete terms.

What’s New in 0.8.14

v0.8.14.0 is the “On Your Clock” release — a fast follow to v0.8.13.0, cut two days later, and the first release deployed to the consolidated host (alpha.pipermorgan.ai has been served by Fly since 2026-09-22 — the one-time re-login you were asked for was that move; your data came across intact).

Your preferences survive restarts (#1574):

• Timezone, reminder settings, working mode, and the calendar-setup offer used to live in server memory and silently reset on every deploy — the “why does it keep asking me?” class. They now persist per user.

Every clock face is yours, and labeled (#1576, #1575, #1577, #1556):

• Standup exports, the agenda, and “what’s my day look like” answers render in your timezone with the zone shown (“3:00 PM PDT”) — never a bare “4:27 PM” and never the server’s UTC clock.
• “Today” and “yesterday” are your calendar days; free-time blocks are computed on your clock. After 5pm Pacific you no longer get tomorrow’s day.
• The agenda had been printing “TBD” for everyone with a connected calendar — a key mismatch, not a missing meeting. Fixed, and the “Focus Time Available” section is reachable again.

Add a project in one line (#1856):

• add project One Job with repo Design-in-Product/one-job creates and links the project in one turn. Leave the name out and Piper tells you the exact line to type and how to cancel — never the same canned question twice.

Honest outcomes on GitHub writes (#1858):

• Closing an issue that doesn’t exist now says “There’s no issue #99999 in owner/repo — nothing was changed” instead of “it may or may not have gone through”. The uncertain wording is reserved for genuinely unverifiable cases.

Fixes and honesty, continued:

• The preferences pages work again (#1864) — every /api/v1/preferences/* route had errored since its auth dependency landed.
• One “Add to Slack” path instead of four (#1499).
• One valid key from any provider is enough to start, and a bad key gets one of four specific explanations (#1823, #1824 — shipped in 0.8.13, first deployed here).
• Counts tell the truth: a list that hit the API page cap reads “100+”, never an invented exact number (#1778).

See Release Notes v0.8.14.0 for full details, and releases/README.md for prior release history (v0.8.13.0’s “Nothing Invented, Nothing Borrowed” notes cover the standup-interview rebuild this release inherits).

Chapter 2: Testing

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

What to Test in 0.8.14

Time and persistence are the focus — every clock face was audited and 27 of 50 were wrong or unlabeled. Is every time on YOUR clock? Do your settings stay set? Does the agenda show real times? (0.8.13’s standup walks still apply — they’re in releases/README.md’s history if you want to re-run them.)

Test Walk 1: Your timezone, everywhere (#1576, #1577)

1. Under Settings → Preferences, set your timezone (pick one that is NOT UTC)
2. Ask “what time is it for me?” — check the answer names your zone (“PDT”, not bare)
3. Export or view a standup — check every timestamp carries a zone label and matches your clock
4. If you can, test after 5pm Pacific: ask “what’s on today?” — check “today” is YOUR date, not tomorrow’s (UTC has already rolled over)

Test Walk 2: Preferences persist (#1574)

1. Change a preference (timezone, reminder time, working mode) and dismiss the calendar-setup offer if it appears
2. Log out, log back in — still set?
3. Come back after the next deploy (or tomorrow) — still set? The offer stays dismissed? If anything reset, that’s the bug this release claims to have fixed — report exactly which setting.

Test Walk 3: Agenda shows real times (#1576, #1575) — needs a connected Google Calendar

1. Connect Google Calendar under Settings → Integrations (note: the connect flow itself still redirects via the old host, #1852 — if it fails, report it and skip this walk)
2. Ask “what’s my agenda today?”
3. Check: meetings show real times in your zone — never “TBD”
4. Check: a “Focus Time Available” section appears with free blocks that end at YOUR end of day (18:00 in your zone), not the server’s

Test Walk 4: Add a project in one line (#1856)

1. Say add project Test Project with repo <your-github-user>/<a-repo-you-own>
2. Check: one turn — created and linked, with a confirmation naming both
3. Say add project with repo <owner>/<repo> (no name) — check Piper gives you the exact line to type and how to cancel, and doesn’t ask the same canned question twice
4. ⚠️ Known: if Piper asks “Want me to add it now?” and your “yes” goes nowhere, that’s #1855 — use the imperative one-liner instead and report the wording you saw

Test Walk 5: Honest GitHub writes (#1858)

1. With GitHub connected: “close issue 99999 in /" (a number that doesn't exist)
2. Check: “There’s no issue #99999 in owner/repo — nothing was changed” — not “may or may not have gone through”, and nothing else in the repo touched
3. Ask a question-shaped one — “did issue 42 get closed?” — check it READS, never closes (#1794)

Test Walk 6: Honest key errors (#1823, #1824)

1. Paste an invalid key in Settings → LLM Keys, send a chat message
2. Check: one specific sentence about the key/provider — not “Something unexpected happened” (if you see the generic message, report the exact wording)
3. Restore your real key and confirm chat works again — with only ONE provider’s key set, chat should still work (#1823)

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://alpha.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

Covered by Test Walks 1–2 above — that flow was rebuilt this release and is the single most valuable thing to exercise. Also worth trying:

• Abandon a standup mid-conversation (“not now”) and confirm your next message routes normally
• Ask “are we done with that standup?” mid-flow — you should get an honest status answer, not a surprise finalization

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; email xian@pipermorgan.ai for a fresh one.

Forgot your password? Email xian@pipermorgan.ai — there’s no self-serve reset in the alpha yet.

Can’t reach the site? Check https://alpha.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 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-, OpenAI keys sk-.

“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 uses GPT-4-class/Claude models. 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

• Bugs and crashes (with the exact error message)
• Anything Piper said that turned out to be untrue — false claims are our top-priority bug class, even when they’re polite
• Confusing flows, missing features you expected, performance problems
• Successful workflows that delighted you (genuinely useful signal)

How to Report

1. Private: email alpha@pipermorgan.ai. It goes to the alpha team and stays there 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

• Your LLM API keys are stored encrypted (AES-256-GCM field encryption) on the alpha instance and used only to make your own requests; they are never shown back or shared
• Your files, lists, todos, and projects are private by default; sharing requires an explicit grant from you
• Error logs may be reviewed by the team to fix bugs
• This is a small, curated alpha — the operators can technically access the instance’s database; don’t put anything sensitive in it (see the disclaimers above)
• Note: data is not yet fully encrypted at rest (see ALPHA_KNOWN_ISSUES.md)

Questions?

This is alpha software (version 0.8.14.0) — expect rough edges, and thank you for being an early adopter. 🚀

See Also

• ALPHA_QUICKSTART.md — the 2-minute version of Chapter 1
• ALPHA_KNOWN_ISSUES.md — current bugs and limitations
• ALPHA_AGREEMENT_v2.md — terms and conditions
• VERSION_NUMBERING.md — what 0.8.14.0 means
• CONTRIBUTING.md — running the code locally (engineers)

Last updated: September 23, 2026 Software version: 0.8.14.0