OrgFlow

OrgFlow is the official and permanent product name. Its exact capitalization is OrgFlow. Never rename it without explicit approval from the project owner. Every browser tab must display exactly OrgFlow.

Authoritative brand configuration: backend/branding.ts defines PRODUCT_NAME = "OrgFlow", DOCUMENT_TITLE = "OrgFlow", and PRODUCT_DOMAIN = "orgflow.dev". Preserve technical identifiers, integration URLs, storage paths, the iridescent knot and provider icons.

Your nonprofit's knowledge shouldn't leave when your employees do. OrgFlow is an AI-powered institutional memory platform for nonprofits: it turns scattered emails, documents and employee knowledge into living, human-approved playbooks that survive staff turnover. Think Process Street plus AI knowledge management, built for organizations with 10-200 staff.

Links

Website	https://orgflow.dev
Web app	https://orgflow.dev/a/ (sign up, or choose the demo workspace)
API docs	https://orgflow.dev/docs (Swagger UI), https://orgflow.dev/openapi.json
Health check	https://orgflow.dev/health
iOS app	app/iOS/OrgFlow.xcodeproj (talks to https://orgflow.dev by default)

Status: PostgreSQL/Prisma production cutover is deployed and verified; AI paths partly verified live; nothing verified on a physical iPhone. Details: docs/SHARED_MEMORY.md.

Capabilities

Capability	What it does	Where
Connect	Google Drive and Gmail OAuth (read-only); consented text indexing of recent Docs and emails. Microsoft 365 is not built.	Web (Files: connections and indexing); backend/google.ts
Capture	Discover suggests recurring workflows from indexed content and drafts playbooks; voice interview on web (ElevenLabs) and on iOS (on-device transcription); drafts need human approval; runnable checklists; owners and periodic review ("due for review").	Web, iOS (Capture, Library); backend/operations.ts, agent.ts
Ask	Answers grounded in approved articles plus consented email/document excerpts (BM25 retrieval), labeled by source type.	Web agent, iOS Ask; backend/agent.ts, retrieval.ts
Transfer	People records, mark leaving, AI handoff guide from owned playbooks, sent emails and notes, successor onboarding checklist.	Web (People, Knowledge map), iOS Handoff; backend/operations.ts

Platform: organizations with isolation (new org at sign-up or invite code; default shared demo workspace, "Center Valley Pantry", with labeled samples), admin/member roles, audit log, billing through the AWS Monetization sandbox (enforcement off).

Quick start

Requires Bun and Docker Desktop (running) for the local database.

bun run install:app   # installs backend and frontend dependencies; rerun after pulling new backend packages
(cd backend && docker compose up -d)  # local PostgreSQL + pgvector
cp backend/.env.example backend/.env   # first time only; fill values locally, never commit
cd backend && bun run prisma:migrate:deploy && cd ..  # create the local schema
bun run dev           # backend :3003 and frontend :3002

• Web app: http://localhost:3002 (production: https://orgflow.dev/a/)
• Landing page: cd landing && bun install && bun run dev, then http://localhost:3001 (production: https://orgflow.dev)
• Swagger UI: http://localhost:3003/docs; OpenAPI: /openapi.json; health: /health
• Verify: bun run typecheck, bun run test, bun run build
• iOS: open app/iOS/OrgFlow.xcodeproj in Xcode, run on a simulator or device (deployment target iOS 27). It talks to OrgFlowBaseURL in app/iOS/OrgFlow/Info.plist (default https://orgflow.dev); point it at your own backend to test local changes.

bun run dev fails with "DATABASE_URL is required" when the database is not running or DATABASE_URL is not set, and with "Cannot find package" when dependencies are out of date (rerun bun run install:app). To stop the database: (cd backend && docker compose down); add -v to delete its data.

Without LUMINAL_API_KEY, Ask, drafting, Discover and handoff return an AI_NOT_CONFIGURED error. PostgreSQL is required for all app runs; local development uses the pgvector Docker image described above. Pushing to main deploys production (.github/workflows/deploy.yml, deploy/).

Demo walkthrough

Live AI steps need LUMINAL_API_KEY on the backend; Google steps need a Google test user (the OAuth app is in Testing). Use synthetic or consented content only. Live AI steps for Discover and handoff were not verified end to end (provider returned 503 during testing).

The shared demo workspace follows the Center Valley Pantry story: Maria Reyes, Volunteer Coordinator, leaves on October 31, and her playbooks, documents and emails are already indexed and owned by her. Sam Rivera (Operations Manager) supervises; Priya Patel owns donor stewardship.

1. Connect. Sign up with an organization name (or use the demo workspace). On Files, under Connected tools, click Connect for Google and consent as a test user.
2. Index. On Files, in the indexing panel, read the consent text, and index recent Docs and emails. Show the indexed source list and that any source can be removed.
3. Discover. In Knowledge Base, run Discover (suggested playbooks). It proposes recurring workflows (for example donor stewardship) with evidence from the indexed emails and documents.
4. Draft and approve. Draft a playbook from a suggestion, review the cited sources, edit, then Approve and publish. Assign an owner; show the review date and "due for review".
5. Run a checklist. Start a run from the published playbook, assign it, and tick off steps.
6. Ask. Ask a question in the agent. Show the answer citing the approved playbook and email/document excerpts, each labeled by source type.
7. Transfer. In People, mark a person as leaving and generate their handoff guide (from owned playbooks, sent emails and notes), approve it, then start the successor's onboarding checklist.
8. Knowledge map. See playbooks and indexed source material grouped by owner (List or Clusters view), drag an item onto a person to reassign it or drop one person onto another to hand everything over, and open a person's handoff flow. Assigned sources feed that person's handoff draft.

iOS moment: on an iPhone, open Capture (or Handoff), narrate a process or role aloud (on-device transcription), and show the resulting draft in Library awaiting "Approve & publish". Not verified on a physical device.

Docs

• AGENTS.md: instructions for coding agents
• docs/SHARED_MEMORY.md: decisions, current state, handoff
• docs/OPEN_QUESTIONS.md: questions needing a human answer
• docs/MVP.md: original brief; docs/PRODUCT_DESIGN.md: design brief
• docs/GOOGLE_SETUP.md, docs/LUMINAL_FLASH.md
• backend/README.md, frontend/README.md

landing/ is the marketing site at https://orgflow.dev (animated thread-map hero and product walkthrough; links to the web app). Secrets stay in Git-ignored backend/.env and GitHub Actions secrets.

Fictional OrgFlow showcase library (opt-in)

The demo-content/ package contains complete, nonidentifying nonprofit documents, PDFs, CSVs, email examples, a calendar and a verified 90-second ad handoff. It is not a customer dataset.

An admin of a newly registered Center Valley Community Pantry (OrgFlow Showcase) organization can visit Organization and choose Load showcase files and staff. This imports linked original documents and searchable synthetic source text, plus role ownership, reporting lines, successors and draft playbooks. It never touches the shared demo or another organization's data and does not connect a Google account.

Google Drive and Gmail integration require separate OAuth authorization and indexing consent. Instructions: Website import and access.