Whyline — Company Memory Engine

Your company forgot why it decided everything.

The Slack thread from March nobody can find.
The pricing rule nobody can explain.
The vendor you rejected — for reasons that left with Sarah.

Every quarter, someone asks "wait, why do we do it this way?"
Every quarter, you pay six people for two weeks to re-discover what you already decided.

That's not a documentation problem. That's institutional amnesia.

---

Whyline is the why layer that actually ships.

Server-side SQLite your whole team shares — not one browser tab's memory.
BM25 retrieval — not your entire corpus rammed into every prompt.
Receipts on every answer — Slack permalinks, doc links, Jira keys.

Ask: "Why don't we support annual billing in Germany?"
Get the trail. Not keyword soup. The actual reasoning.

Capture from Slack (📌 or /whyline), email (decisions@), meeting transcripts, Notion/Confluence/GDocs, GitHub PRs, Teams, Linear, Jira — or just talk to it through MCP in Cursor.

Clone it. Plug your API key. Running in 60 seconds.
Free. Local. Yours.

Ask why · Capture · MCP + Cursor

---

What changed in v2

We stopped pretending IndexedDB was "company memory." v2 is boring infrastructure that works:

Layer	What it does
SQLite (data/whyline.db)	Shared decision store — survives browser clears, works for the whole team
BM25 retrieval	Top-k decisions per question — scales past a few hundred entries
Provenance	Every decision links back — Slack, email, doc URL, Jira key
Thread-level capture	Full Slack threads, not per-message garbage
MCP	whyline_ask / whyline_extract / whyline_search — Cursor becomes a client

.env is never served. Default bind: 127.0.0.1. Webhook secrets required, not optional.

---

Quick start

git clone https://github.com/Sunradiance/whyline.git
cd whyline
cp .env.example .env
# Edit .env — add LLM_API_KEY
npm run setup
npm run dev

Open http://127.0.0.1:8793

npm run dev uses Flask's built-in server — fine for local use. For a shared team deployment, use a production WSGI server (atomic capture claims make multi-worker safe):

pip install gunicorn
gunicorn -w 4 -b 127.0.0.1:8793 'app:create_app()' --chdir backend

On Windows: pip install waitress then waitress-serve --listen=127.0.0.1:8793 --call app:create_app from backend/.

Minimum .env

LLM_API_KEY=gsk_...
LLM_BASE_URL=https://api.groq.com/openai/v1
LLM_MODEL_NAME=qwen/qwen3-32b
WHYLINE_HOST=127.0.0.1

---

MCP + Cursor

Whyline is an MCP server. Every AI assistant in your IDE can query institutional memory — and write to it.

Tools: whyline_ask · whyline_extract · whyline_search

One-time setup

npm run setup   # if you haven't already

Copy .cursor/mcp.json.example → .cursor/mcp.json (or use the included project config) and adjust the path.

Restart Cursor → Settings → MCP → enable whyline.

Now ask in chat: "Whyline: why did we reject Acme CDN?" — it hits your local SQLite corpus with receipts.

Details: integrations/mcp/README.md

---

Features

• Ask Why — retrieval-backed answers with validated decisionIds + clickable receipts
• Decision registry — search, filter, supersede lifecycle
• Capture — transcripts, email, docs (Notion/Confluence/GDocs), Slack/Teams threads
• Slack — /whyline slash command + 📌 reaction → full thread capture
• decisions@ — CC email alias → extract + persist (setup)
• Jira / GitHub / Linear / Teams / Salesforce — webhook ingest with provenance
• Memory Brief — export for leadership

Demo seed data: DACH billing, Acme CDN, Teams feature kill.

---

Integrations (live)

Source	Endpoint	Auth
Email decisions@	POST /api/integrations/email/ingest	X-Whyline-Secret
Slack /whyline + 📌	/api/integrations/slack/commands + /events	Slack signature
Meeting transcript	POST /api/integrations/transcript/ingest	session / API key
Notion / Confluence / GDocs	POST /api/integrations/doc/ingest	session / API key
GitHub / GitLab	POST /api/integrations/github/webhook	X-Hub-Signature-256
MS Teams	POST /api/integrations/teams/ingest	X-Whyline-Secret
Linear	POST /api/integrations/linear/webhook	X-Whyline-Secret
Jira	POST /api/integrations/atlassian/jira	X-Whyline-Secret
Salesforce	POST /api/integrations/salesforce/webhook	X-Whyline-Secret
MCP	npm run mcp	local stdio

Docs: integrations/email · integrations/slack · integrations/mcp · integrations/github

---

Team mode & access control

Set WHYLINE_AUTH_MODE=team (or open) for multi-tenant deployments. Solo stays the zero-config default.

Role	Read	Create decisions	Manage decisions	Invite/remove members, mint/revoke tokens	Workspace settings
viewer	✓
member	✓	✓
admin	✓	✓	✓	✓
owner	✓	✓	✓	✓	✓

Deprovisioning: remove a departed user's membership (DELETE /api/workspaces/<id>/members/<user_id>) and audit workspace tokens (GET /api/workspaces/<id>/tokens), revoking any that person may have seen (DELETE /api/workspaces/<id>/tokens/<token_id>).

Service tokens are infrastructure credentials — workspace-scoped, not user-scoped. MCP, CI, and integration keys stay valid until explicitly revoked. That is intentional: tokens outlive any single human account. On offboarding, owners rotate tokens via the list endpoint; member removal does not cascade to tokens they minted.

Auth mode ratchet: a DB that has booted team/open refuses to silently downgrade to solo if WHYLINE_AUTH_MODE is unset. Override only with WHYLINE_ALLOW_SOLO_DOWNGRADE=1.

---

Tests

npm run test

97 tests · CI on push to main

---

MIT licensed. No paywall. Optional voluntary tips: SUPPORT.md (off by default).

---

Whyline — decisions remembered.
The why layer your company never had — now with receipts.