Real World Use Cases
Real World Use Cases
Every scenario below is real. Pulled from actual message logs and daily notes across two weeks of a family using Lobster. No demos. No mockups. Just life, handled.
”Tell Omar the kids want to see a movie”
Your spouse is at the resort. You’re not checking your phone. She texts Lobster:
“Hi Lobster Can you text Omar to let him know that the kids want to see a movie today Send Help 12:30 showing at the theater nearby”
Lobster relays it to you via iMessage. You respond. She follows up:
“What time do you recommend we leave the resort to park, get refreshments to make a 12:30 movie”
Lobster searches drive times, adds a buffer for parking and snacks, and replies: leave by 11:45 AM. No app switching. No Googling. Just an answer.
”Where did we travel last year?”
Your daughter drops a question in the family group chat:
“Lobster do you know where we as a family traveled in [country] last year”
Lobster checks the family’s Travel Hub, finds the trip — dates, duration, number of nights. She follows up from memory:
“I know we stayed at a nice hotel on the water near a famous landmark”
Lobster identifies the hotel. Now the whole family is reminiscing in the group chat, with Lobster filling in the details nobody can quite remember.
She also asks:
“What are our next two trips? Where are we staying?”
Lobster pulls up the itinerary. Trip planning feels less like work and more like anticipation.
”Add this to the calendar”
“Hi lobster! Is there any chance you can add [event] to the calendar?”
Done. Event created on the shared Family calendar. No login required. No app needed. Just a text.
Your flight just landed
You didn’t ask for this. Lobster was already tracking the flight via FlightRadar24. When the wheels touch down, you get:
- A landing notification
- Check-in instructions for the hotel
- Driving directions from the airport
- A reminder created with arrival details
Your spouse gets the same info in the family group chat — driving directions, tipping guides, and what to expect at check-in.
”Play Bad Bunny on Sonos”
“Play Bad Bunny on Main Floor Sonos”
Music starts. Multi-room Sonos control, Spotify and Apple Music, per-room volume defaults. Play, pause, skip — all from a text message.
You never missed that 7 AM meeting
Lobster checks your work calendar every 30 minutes. It knows the patterns:
- Early morning meetings (7-8 AM)? Reminder sent the evening before.
- Late meetings (after 8 PM)? Reminder sent at 5 PM day-of.
You didn’t set these reminders. You didn’t configure rules. Lobster just pays attention.
Forward emails to Lobster
Got an email you need Lobster to handle? Just forward it to his iCloud mailbox. He checks it two to three times a day and knows what to do:
- Only processes emails from contacts and verifies sender authenticity (DKIM + SPF) before trusting anything
- Flags what matters
- Processes travel agent emails into actionable check-in instructions
- Archives what’s done
- Drafts replies
You forward, review, and approve. Lobster does the rest.
One Family, Three Agents
Not everyone in the family needs the same access. Lobster runs as three agents with different permissions:
You get everything — email, calendar, travel, smart home, monitoring, messaging, reminders.
Your spouse and kids get what they need — shared calendars, family travel plans, reminders, contacts, web search — through their own DMs or the family group chat.
What they can’t do: read your email, run system commands, see financial details, or access your work calendar. This isn’t a setting you toggle — it’s enforced at the architecture level. Per-agent permissions, tool-level denials, workspace isolation.
Your daughter texts “Check Omar’s email” and gets a polite refusal: “I don’t have access to Omar’s private email.” Tested. Verified. Secure.
It doesn’t just respond — it watches
Every 30 minutes, Lobster runs a heartbeat:
- Checks for new email worth flagging
- Reviews the family WhatsApp group
- Scans the calendar for upcoming meetings and events
- Monitors messaging infrastructure health
- Self-heals if something goes down
If the messaging bridge crashes at 3 AM, Lobster detects it and restarts the service. You wake up to everything working.
”Where is everyone right now?”
Your family is spread across timezones. You’re in Belgrade, your daughter is on a flight from Istanbul to DC, your wife and son are at home in Seattle. Lobster knows all of this — not because someone told it, but because a background job checks travel itineraries twice a day and writes each person’s current city, country, and timezone to a shared state file.
When your daughter texts Lobster at 2 PM her time, it knows that’s 8 PM yours. When it sends you a meeting reminder, it uses your local time, not the server’s. When someone in the family group chat says “let’s call tonight,” Lobster can flag the timezone gap before anyone books the wrong hour.
No one configured timezone rules. Lobster just keeps track.
Your flight status changed — Lobster already told the right people
Travel Hub sends a webhook when something changes: a flight departs, lands, gets delayed, or a hotel check-in opens. A dedicated travel agent receives these events and decides what to do.
Not every event is worth a message. A routine status update at cruise altitude? Ignored. But when wheels touch down in a new city, the right people get notified:
- You get check-in instructions, driving directions, and a reminder created with arrival details
- Your spouse gets a message in the family group chat with arrival info and logistics
- Nobody gets spammed with “flight is still in the air” updates every 30 minutes
The travel agent classifies each event, decides who cares, and routes the notification. It’s not a dumb webhook forwarder — it’s an intelligent filter between raw travel data and the family’s attention.
”Save this for the trip”
You’re researching restaurants for an upcoming trip. Instead of bookmarking tabs or pasting into a notes app, you text Lobster:
“Save this for the Vienna trip — Steirereck, rated #1 in Austria, reservation needed 4-6 weeks ahead, lunch is more casual than dinner”
Lobster writes it to a trip note in the Obsidian vault. Later, when the trip is approaching, you ask:
“What restaurants do we have saved for Vienna?”
Lobster searches the vault and returns everything. The vault syncs across all your devices via Obsidian Sync, so you can also browse the notes directly in the Obsidian app on your phone.
Daily notes work the same way. Every evening, Lobster reflects on the day and writes a diary entry — events that happened, decisions made, things worth remembering. Months later, you can ask “what did we do on March 8th?” and get an answer.
The mailbox sensor
Your smart mailbox has a sensor. When the door opens, HomeKit fires an event. A dedicated smart home agent receives it, recognizes the pattern, and marks “mail delivered today” in the agent’s state.
Later, during a heartbeat check, Lobster mentions it: “Mail was delivered today.” If you ask “did I get mail?” — it already knows.
This is one of dozens of HomeKit events flowing through the system. Door sensors, power monitoring, temperature changes — a dedicated agent classifies each one, discards the noise, and only surfaces what matters. You don’t get a notification every time a door opens. You get one when something is unusual.
”What should I watch tonight?”
You’re staring at the TV with nothing queued up. You text Lobster:
“What’s on my Trakt watchlist?”
Lobster pulls up your queued shows — filtered by movies or TV if you want. You pick one. Next week, you’re at the Oscars after-party (in your living room) and text:
“Add the Best Picture nominees to my Trakt watchlist”
Done. All of them, in one message. You can also ask what you’ve been watching lately, check if you’ve already seen something, or mark a show as watched — all from a text message. Your Trakt profile stays up to date without ever opening the app.
”Get me a table at Carbone”
You’re trying to eat somewhere good on Friday. You text Lobster:
“Find a 7:30 table for 4 at Carbone on Friday”
Lobster searches Resy, checks availability, and comes back with the open slots. You pick one. Lobster books it — right from the text thread. Confirmation drops into the conversation; the reservation shows up in your Resy account; you get the usual confirmation email from the restaurant.
For the hard ones — the places that release tables at 10 AM exactly 30 days out and vanish in seconds — you skip the refresh-at-10 ritual entirely:
“Snipe a 7:30 for 2 at Don Angie, May 15th, the moment the window opens”
Lobster queues the job. On release morning, the scheduler fires, the booking goes through, and you get a text that it’s confirmed. If the slot’s already gone it comes back with the closest alternatives and asks whether to grab one.
OpenTable works a little differently — no public API means no auto-book — so Lobster sends a pre-filled deep link and you tap Confirm on your phone. One tap, no form filling, no copying restaurant names across apps. Tock and SevenRooms round out the coverage for the restaurants the others don’t handle.
Listing, canceling, and changing your mind all happen from iMessage too. No app switching. No logging in. Just texts.
”Dump this article”
You’re reading something interesting — a blog post on AI agent architecture, a podcast episode featuring someone you follow, a gist from a researcher you admire. Instead of bookmarking it and forgetting:
Lobster extracts the full content, saves an immutable copy to the knowledge base, creates wiki pages distilling the key ideas, and cross-links them with everything you’ve captured before. The article on agent architecture connects to last week’s podcast episode on the same topic, which connects to the design notes you dumped a month ago.
For podcasts, it goes deeper — downloading the audio, transcribing the full episode via Whisper API (or pulling YouTube captions for free), and synthesizing the transcript into wiki pages. A 4-hour interview becomes a structured knowledge page with key quotes, arguments, and cross-references to related concepts.
Over time, the knowledge base becomes a personal wiki maintained by the agent — not a graveyard of bookmarks, but a living, interlinked map of ideas. Ask Lobster a question about any topic you’ve captured, and it synthesizes an answer from all your sources with citations.
A daily podcast made from the newsletters you didn’t read
You subscribe to too many newsletters. Stratechery, Benedict Evans, MacStories, three Substacks you forgot you signed up for. They pile up in a Newsletter folder in Instapaper and you’ll never read them.
Every night at 1 AM, Lobster turns whatever landed there that day into a podcast.
It pulls each new article, runs the body through OpenAI’s TTS in Nova’s voice, generates a per-article cover image, and uploads one episode per article to a private Spotify show called Omar’s Instapaper Daily. Each episode is tagged with the publication it came from — “Source: Stratechery”, “Source: Benedict Evans”, “Source: The Verge” — derived from the article’s URL, the email’s HTML header, or the first non-tracker outbound link, with the original article linked when it’s public. Articles that have already been turned into episodes are remembered, so you never get the same one twice.
In the morning you open Spotify, pick the car, and the newsletter backlog plays on the commute. The publication name on each episode tells you what you’re about to hear before it starts. The reading you would have done — but didn’t — happens anyway.
”How’d your daughter get into stained glass?”
Every morning at 9 AM, you get one question from Lobster.
Not “what’s your favorite color?” Something specific. Something it could only ask because it’s been reading your notes:
“You mentioned your daughter is into stained glass. How’d she get into that?”
You answer when you get a moment — usually 30 seconds while you’re making coffee. The next morning, Lobster has filed the answer to the right place and is asking a new question.
The cron runs daily. The agent reads existing memory files, finds a gap, and writes one thoughtful question to your DM. You reply whenever. The next day’s run processes yesterday’s answer, updates the right file, and surfaces a new gap.
One question. One answer. Filed to the right place. Every day.
After six weeks of this drip, the context is richer than any 30-minute onboarding interview could produce. It catches the stuff you’d never think to volunteer: morning routines, how you take your coffee, that your spouse is going back to school, that you played guitar in bands growing up, that you lived in Denver for five years and that’s why you root for the Avalanche.
The agent doesn’t have to guess what to ask. The memory files are full of half-finished context — a name dropped without backstory, a hobby mentioned once, a city referenced in passing. Each morning it picks one and pulls the thread.
Monthly curated digests in your inbox
Every first of the month, you and your spouse get an HTML email from the agent — scannable, organized by category, one line per show with a clickable link to the ticket page:
Concerts (88) Tue May 12 · Florence + The Machine · Climate Pledge Arena Wed May 13 · Demi Lovato · Climate Pledge Arena May 13–24 · The Phantom of the Opera (Touring) · Paramount Theatre · 16× …
Comedy (5) Sat May 09 · Zakir Khan Live · Climate Pledge Arena Sat Jun 13 · Chelsea Handler · Paramount Theatre …
Under the hood, a skill queries the Ticketmaster Discovery API for your watchlist of venues (the Paramount, Moore, Climate Pledge, Benaroya, WAMU, McCaw, 5th Avenue, the Showbox, the Gorge, Marymoor, Chateau Ste. Michelle, Tacoma Dome, Lumen for Seahawks games only) and renders the next 60 days grouped by category. Long-running residencies — Phantom’s 16 performances, Seattle Opera’s Carmen run — collapse to a single line with a count, so the list stays scannable rather than drowning in repeats.
A monthly cron triggers the agent, which runs a zero-argument wrapper script that generates the HTML and sends via Apple Mail using AppleScript’s html content property (the one way to get rich-text email through Mail.app programmatically). No need to open Ticketmaster every month, no missed touring acts.
What Lobster handles most
Real usage patterns from a real family, ranked:
- Message relay — “tell Omar,” “let the kids know,” “forward this to…”
- Calendar — “what’s this week look like?”
- Travel — trip details, tipping guides, hotel lookups, real-time flight tracking
- Logistics — drive times, departure planning, practical questions
- Proactive alerts — meeting reminders, flight landings, smart home events
- Notes & research — trip journals, restaurant saves, reference lookups from the vault
Every family member finds their own rhythm. The spouse leans on relay and logistics. The daughter asks about trips and adds calendar events. The owner uses everything — email, music, monitoring, notes, the works.
The full toolkit
Communication
iMessage, WhatsApp monitoring, email — Lobster reads, writes, relays, and summarizes across all of them. Multilingual. Culturally aware.
Calendar & Scheduling
Shared iCloud calendars for the family. Smart meeting reminders that learn your patterns. Create events from a text message.
Travel
A family travel hub with itineraries, hotel details, destination guides, and tipping etiquette. Real-time flight tracking with maps and landing notifications.
Smart Home
Multi-room Sonos, Spotify, Apple Music. HomeKit sensor monitoring — mailbox, doors, power, temperature. Play, pause, volume — by room, by voice, by text. Unusual events surfaced; routine ones silently logged.
Notes & Knowledge Base
An Obsidian vault shared between the family and the agent. Daily diary entries, trip journals, restaurant research, airport notes, packing lists. A dump command captures any URL, podcast, or text into an AI-maintained wiki — extracting content, creating cross-linked wiki pages, and transcribing full podcast episodes. The agent writes to it, the family reads it on any device via Obsidian Sync. Searchable from a text message.
Dining
Restaurant reservations across Resy, OpenTable, Tock, and SevenRooms — search, check availability, book, cancel, and snipe hard-to-get tables the moment the booking window opens. All from a text message. Destructive actions confirm before firing; OpenTable uses a one-tap deep-link hand-off since no public API exists.
Entertainment
Trakt.tv integration for tracking movies and TV shows. View watch history, check your watchlist, search for titles, and mark things as watched — all from a text message. Cross-reference with Obsidian notes to backfill years of viewing history.
Memory
Lobster remembers. Family preferences, friend groups, past trips, lessons learned. It builds a knowledge base over time so it gets better at helping your family specifically.
Content
Image generation, text-to-speech, presentations. When you need something created, not just retrieved.