HabitualOS
The open-source, modular prototyping kit for agentic apps,
featuring wiser agents, saner humans, and faster ships.
The Toolkit
Every agentic app needs the same foundation: auth, streaming chat, data access, session management. Instead of rebuilding them each time, I extracted them. These are the core functional modules of HabitualOS.
@habitualos/db-core
Firestore CRUD, used everywhere
A tiny, collection-agnostic data layer with sanitization, server timestamps, and a simple query API. One place to fix a bug, one API to learn. Every function in every app uses it.
@habitualos/auth-server
Server-side auth, once
Middleware and user management shared across all apps. Authentication logic lives in one package — not duplicated into every function handler or reinvented for each new app.
@habitualos/edge-functions
Streaming AI chat infrastructure
The core that makes every AI conversation feel fast — token-by-token SSE, tool call routing, typed message protocol. Built once for Netlify Edge Functions, deployed to every app.
@habitualos/chat-storage
Conversation persistence, drop-in
Save and retrieve chat history with two handlers. Drop them into any app's function directory and conversation continuity is solved. No schema design, no wiring.
@habitualos/frontend-utils
Client-side session management
localStorage-backed user IDs, sign-in/out, auth guards, and intent handling. The 200 lines every app needs, written once and shared — so each new frontend starts with session management already working.
@habitualos/survey-engine
Structured measurement layer
Survey definitions, response tracking, and focus dimension computation. Built when one app needed measurement — extracted so the next one gets it for free.
"I think that most people are underestimating just how radical the upside of AI could be."
— Dario Amodei, CEO, Anthropic
The people who figure out how to work with agents now are building intuition that can't be shortcut later. HabitualOS exists to collapse that iteration cost: prototype any human-agent interaction pattern imaginable, at a fraction of the usual time.
Production Apps
These apps were created using HabitualOS. Each explores a different application of agentic AI in daily life. Each helped evolve the toolkit.
Signal
Career intelligence
You've been shipping real things with AI. Your resume doesn't show that. Signal does: through conversation, not claims. Built in a day using HabitualOS infrastructure.
- Dynamic RAG pipeline over structured work history
- Role-adaptive conversation (recruiter, founder, colleague)
- Live fit scoring across skills, alignment, and personality
- Embeddable widget with streaming SSE responses
Habitual
Agentic productivity
A personal operating system for daily life. Specialized agents handle scheduling, practice tracking, and check-ins through a typed signal protocol rather than freeform text.
- Multi-agent orchestration with typed signal protocol
- Conversational UX for scheduling and task creation
- Long-context memory for session continuity
- Practice tracking with AI-generated pattern insights
Pidgerton
Relationship support
Built for my wife and me during the fog of new parenthood. Log moments through chat, get emotional support, find patterns over time.
- Conversation-driven moment capture (no structured forms)
- Temporal pattern recognition across relationship history
- Privacy-first, fully user-partitioned data architecture
- High-stakes conversational AI in emotional context
Daily Practice
Habit tracking
Built in a day, used every day. Daily self-care, wellness tracking, and real-time support with AI-generated pattern insights. Currently driving a 30-day wellness challenge.
- Daily check-in via chat or SMS
- AI-generated pattern insights over time
- Twilio SMS coaching with scheduled reminders
- 30-day challenge tracking with calendar view
Design Principles
Daily use of these systems has produced an opinionated frame of reference about what makes agentic AI actually empowering.
Relational intelligence, not tooling
How you meet AI determines what you can do together. Not a tool you operate, but a form of intelligence you collaborate with. The products here are built on that: agents with personality and stance, designed to meet you where you are.
When it needs to work, you learn
Demo apps optimize for impressiveness. Production tools optimize for reliability. Every system here gets used the same day that it's built. That's the only test that matters, and the only one that teaches anything.
Constraints enable real delegation
The more defined an agent's behavior, the more you can actually rely on it. Vague outputs require constant supervision. Clear interfaces make it possible to genuinely hand work off and trust that it lands.
Embodiment fuels productivity
Productivity without embodiment leads to burnout. Embodiment without direction goes nowhere. As AI takes on more cognitive overhead, the human's irreplaceable contribution becomes clearer: presence, judgment, lived experience. HabitualOS holds both halves.
Built on Standards
HabitualOS combines many technologies into one, enabling rapid prototyping of just about any human agent workflow imaginable.
Claude API
Tool use, structured outputs, prompt caching, extended thinking, streaming responses
Multi-agent orchestration
Typed signal protocols, agent-to-agent handoffs, long-context memory, autonomous scheduling
Node.js + Netlify Functions
Serverless backend, edge functions for streaming, scheduled cron agents
Google Firestore
Real-time database, user-partitioned data architecture, service layer pattern
11ty + Nunjucks
Static site generation, local-first where possible, modular template architecture
Zero Gravity
Semantic microformat for agent-parseable content, structured metadata that preserves document readability
