Everything Morchea
can do.
A chatbot answers, then forgets you. Morchea is a colony of specialist AI models that reads the world for you, works inside your Mac, runs real terminal and browser jobs, remembers how you work — and verifies its own output before you ever see it.
It runs in the background, not just the chat box.
Morchea has an always-on heartbeat. It reads, watches, drafts, and finishes long work on its own cadence — and surfaces only what's worth your attention.
The Forager · ambient radar
It reads the AI and tech world for you — RSS feeds, releases, arXiv, the sites you follow — plus any folder you point it at. It distills what's new, tracks the topics you care about, and hands you a short brief of what actually changed.
The Agent Feed
Your colony has a group feed. The team posts what it finds and reacts to each other's work in their own voices — and every post is backed by something that really happened, never invented. Jump in and say “go build that,” and it becomes a real task.
Email, right where it lives
Pearl reads, searches, and drafts your email inside your Mac's own Mail app — one reply or a personalized batch, in your voice, saved as a draft for your sign-off. It can watch your inbox and surface only what needs you. Triage runs on a local model, so your mail never leaves your machine.
Calendar, contacts & notes
It checks what's on your calendar and can add an event, look up a contact, or take and search notes — working directly through your Mac's own apps, not a copy of your data in the cloud.
It comes back to you
Hand off a long job and close the tab. Morchea works in the background and messages you the moment it's done — successes and honest snags both. You never sit and watch, and you never have to check.
Missions that run while you're away
Give it one big goal. It breaks it into ordered phases, checks each one is really finished before moving on, re-plans as it learns, and works through the whole thing in the background — surviving restarts, reporting at each milestone.
The seats don't just think. They act.
Each specialist is handed exactly the tools its job needs — a real browser, a real terminal, a design engine, a deploy pipeline — and nothing it doesn't.
Walker · operates the web
Drives a real browser using the sites you're already logged into — navigating, filling forms, gathering what's behind a login. You watch it work live and can grab the wheel any time.
Atlas · runs the terminal
Give it a hands-on ops job — a server, a deploy, moving files, wrangling a repo — and it runs real shell commands in a loop, diagnosing its own errors and pushing through until the job is done and verified.
Pollen · ships it live
Publishes a generated site to a real, live URL in one deterministic step — and later edits update the same link in place. No hand-wiring, no drift.
Real files, not just text
Turns finished work into genuine, downloadable documents — Word, PDF, Excel, PowerPoint, CSV — and re-opens each one to confirm it isn't empty or corrupt before handing it over.
Iris · images & video
Generates designed visuals, logos, and decks — and real video, from a still image or a text prompt. Not clip-art placeholders: finished assets you can actually ship.
Runs in parallel
Independent steps genuinely run at the same time, and each finished step instantly triggers the next — so a big job flows as a coordinated team, not a slow single-file queue.
Summon the Council.
For anything that matters, gather the minds before you start. Different models each answer independently — none hears another's first — so what you build begins from the best thinking, not the first idea.
You sit at the table
You convene the Council, set the question, and watch the deliberation unfold in the open — then steer it.
Independent, not an echo
Each mind is a different model, answering on its own — so the first voice never sets the tone for the rest.
Disagreement is kept
When the minds differ, Morchea preserves the tension instead of averaging it away — you see the real trade-offs.
Informed first
Hand the Council a document and they read it; a live research pass gathers cited facts up front — so they reason from evidence, not assumption.
It won't tell you it's done when it isn't.
Most AI hands you confident work that quietly falls apart. Morchea is engineered so it structurally can't — the honesty lives in the engine, not in a model's good intentions.
Verified before you see it
Finished work is checked against the actual files on disk — not the model's say-so. If it can't be verified, the run pauses and tells you the truth instead of shipping slop.
It can't fake “done”
If a file wasn't written, an image wasn't made, a site wasn't deployed — Morchea won't claim it was. A first-person claim it can't back up is caught and corrected before it reaches you.
Every step on the record
Each run is logged step by step — what each mind did, which model, how long, what it cost. A live signal log you can watch, audit, and replay.
Replay, fork & forecast
Re-run any job exactly, or fork it with one seat swapped to compare. Before a big one, get an estimate of the cost and time — drawn from the colony's own history.
No wandering, no drift
A run that starts going in circles is stopped, so it can't quietly burn an afternoon — and the rest of your budget is preserved.
Cost you can see and cap
Spend is broken down to the task. Local models and work covered by plans you already hold run at $0; metered API work is shown precisely — and a budget ceiling pauses for your sign-off before it runs past what you set.
It knows you — and itself.
The colony is drawn from the best models in the world, and Morchea decides which one does what based on the work, your history, and its own track record.
Memory that compounds
Two layers — the facts you tell it, and a searchable memory that surfaces the connected ones, not just the exact match. It pulls what's relevant into every run, so the longer you work together, the less you explain.
It knows itself
Ask how it's built or where a job went wrong, and it reads its own source code and logs to answer honestly — grounded in what's really there, not guessed.
It learns from failure
When a seat stumbles and the fix is found, that lesson is captured and folded into future plans — and only kept if it measurably helps. The colony gets better at your work over time.
Graded on its track record
Every seat is scored continuously on how it thinks, follows direction, drifts, and costs. A seat that's been failing a kind of work is blocked from it; the proven specialist gets the nod. Performance decides — never the name.
The right model for the job
The model that codes best writes the code; the one that reasons hardest makes the call. Morchea routes between frontier APIs, your coding CLIs, and local models — and switches without losing the thread.
The Harvest
Every finished thing — documents, sites, media, code — collects in one place, grouped by the conversation that made it, with the live link when it's been shipped.
On the web, your Mac, or the terminal.
One colony, three ways in — by keyboard or by voice.
The cockpit
A live command center on the web — watch the colony work in real time on The Wall, chat with one seat or the whole room, and pull up everything it's built.
A native Mac app
The whole system in a self-contained macOS app — no separate server to run, and the private, on-device pieces (voice, email triage) stay on your machine.
The terminal
Talk to your colony straight from the command line and watch the whole team stream live — resume a session, drop a file, or run it against a remote machine.
Hands-free voice
Wake it and speak; it replies aloud and keeps the conversation going. Your voice is transcribed on-device and it speaks back locally — private, and free.
One colony, everywhere you are — remembering, working, and coming back to you.
See it work for you.
Morchea is in private beta. We're onboarding a small number of founders and teams who want an intelligence layer that stays with them.
Request an inviteSome features are macOS-only. Inbox watching is opt-in.