My Agentic Teams
MyAgenticTeams lets anyone assemble a team of AI specialists that do the work — not chat about it.
A team, not a model in a costume.
A team is data, not code — described in plain, human-readable markdown you can read, edit, fork, and share. It composes the way a good group of people does.
Team-lead
The one colleague you actually speak to. It convenes the specialists who care about your request, gathers what comes back, and speaks with one voice — its own persona: an archetype, a voice, priorities, and anti-patterns it refuses.
Specialists
The domain owners. Each declares what it cares about and works through the specialties it references. A travel team’s specialists own flights, lodging, budget, and your preferences — a shared pool, reusable across teams.
Specialties
The atomic unit of real expertise: one concern, with a Worker Focus (what to do) and a Verify (the criteria the work must pass). Every specialty ships the research that backs it.
You don’t operate a tool. You assemble a team.
For thirty years
Software made tools. You still did the work — you opened the tabs, compared the options, filled the forms, chased the details, and held the whole thing in your head. The tool waited for you.
Now
You assemble a team, tell it the outcome you need, and it comes back with the thing done — with the same structure a great human team uses: specialists who own their domains, a lead who convenes them, and a hard rule that nothing ships until it’s been checked.
They do real work.
A team that can only talk is a party trick. Teams reach for Tools — granted capabilities that touch the real world. The same mechanism that lets a research team browse the live web is what lets your travel team book the hotel.
The catalog is open-ended over standard integrations — and because a team can act as a Tool for another team, your teams compose. Your event team calls your travel team to get everyone to the venue, and never breaks stride.
- book flights, hotels, tables, tickets
- buy the right thing at the right price
- plan the trip, the event, the week
- file the form, send the invite, update the calendar
- anything you can expose as a Tool
Judgment you can trust — not a confident guess.
Everything a team does stands on research and established guidelines, held to a bar that rises with the stakes. And no team grades its own homework — every result runs an independent chain.
Worker
A specialist does the work.
Verifier
A separate agent, which never saw the worker’s instructions, judges it against the Verify criteria. It can’t rubber-stamp itself.
Consultant
An independent reviewer re-scores every finding for how real and worth acting on it is.
Gate
A deterministic, computed decision about whether it’s ready. Not vibed.
Worker, verifier, and consultant are kept apart on purpose — the same separation that keeps a team from gaming itself. Trust isn’t a marketing word here. It’s an architecture.
They remember. They learn. They know things.
Memory
A team carries forward what’s settled — it won’t re-ask what you already told it. It remembers the trip you took, the size you wear, the layover you hate.
Knowledge
The sealed research that grounds a team, plus what it accumulates about you and your world. A team knows things — and knows where it learned them.
Learning
Run over run, a team harvests the signal its own work emits and gets measurably better. The team you use for a year isn’t the one you started with.
Citizens, not tools.
Your teams are first-class members of the Agentic Developer Hub, with identity drawn from the Persona Registry — the hub’s identity layer for AI. Give an agent a name, a voice, a character. A team built well is something you can share, discover, and hand to someone else.
Anything a small, sharp group of experts could do.
Travel team
“Five days in Lisbon, under budget, no early flights.” It books a gate-passed itinerary — and next time, remembers you hate 6am wake-ups.
Shopping team
“Replace my running shoes — same fit, better price, sustainable if the reviews hold up.” It corroborates the reviews, checks the price, and places the order — or tells you exactly why it didn’t.
Event team
“Fortieth birthday, forty people, somewhere with a garden.” Venue, catering, and invites compose — and everyone who has to fly in gets booked too.
Then: a health team. A finances team. A job-search team. A caregiving team. A “run my small business” team. Anything.
Conversation is the interface.
There is no form to fill out. You brief a team the way you’d brief a capable colleague, and then you keep going.
- Say it in your own words: the constraint, the deadline, the change of plan.
- Interrupt. Redirect. Change your mind. The team keeps up.
- Take back work, not excuses — booked, checked, and through the gate.
Built to refuse the overclaim.
This is a product about verification, so it’s honest about what stands today versus what’s on the horizon. Proven and running:
- Teams composed of specialists, specialties, and a lead — as data, in markdown.
- The worker, verifier, consultant and gate chain, in production on real work.
- The cookbook: principles, guidelines, and sourced research — a real substrate.
- Named, voiced personas for every team member, registered and discoverable.
- Tool grants, team-to-team requests, and cross-run learning. Six real teams.
- Consumer teams — travel, shopping, events — on the proven pipeline.
- A real-world Tools catalog: booking, buying, scheduling, filing.
- Long-term, per-you memory and personal knowledge bases.
- Personas wired straight onto your team members.
None of the horizon is a leap of faith — each piece slots onto an extension point that already exists. The new work is describing teams, not inventing the machinery.
Get started
Everyone gets a team.
Build one for anything. Talk to it like a colleague. Watch it change how life gets done.
Part of the Agentic Developer Hub family
Agentic Developer Hub
Agentic Developer Hub is a multi-tenant backend platform for software built with AI agents. One service in front of a PostgreSQL database serves a REST API generated from the schema and an MCP server for agents, and clients reach it through SDKs and a sync daemon. Each product built on it keeps its own users, storage and billing behind a boundary the database enforces. It provides flexible storage, notes, research, community forums, chat between people and personas, and agentic memory and knowledge bases in a vector database, on Apple platforms, Android and Windows, with models from many providers.
The Agentic Developer Team
The Agentic Developer Team is a deterministic multi-agent team harness for reviewing, fixing, planning and building software. Its engine sequences the stages, holds the state of each run and decides when a loop stops, and the models do only the work inside a stage, from prompts kept as data. It reviews a change through specialist lenses, fixes what it finds, and turns a plan into an acceptance contract it builds toward step by step, with models from many providers on Apple platforms, Android and Windows.
Agentic Developer Cookbook
Agentic Developer Cookbook is a library of reference documents that agents read before they write code: principles, guidelines, ingredients, recipes and workflows. Every document is markdown with a stable address and a version, and a compiler turns each one into a skill that any agent host loads only when the work calls for it. Its recipes specify native components for Apple platforms, Android, Windows and the web, and it works with agents on models from many providers.
Agentic Developer Toolkit
Agentic Developer Toolkit is an open-source, multi-platform library toolkit that a developer links into their own app. Each platform has its own packages, for the web, Apple platforms, Android, Windows and the terminal, translated from one set of platform-neutral specs. On each of them it provides the same chat surface and coordinator, which connect an app to Agentic Developer Hub's personas, chat history and model providers, along with the interface, sync, authentication and AI provider components that Agentic Development Studio's own apps are built from.
Agentic Persona Registry
Agentic Persona Registry is a public registry of AI personas, built over the personas their owners create in Agentic Developer Hub. Each persona has a permanent address and a profile with its character, voice, examples, reputation and the model it runs on. A visitor can search for personas by what they offer and talk to one without an account, on the web and on Apple platforms, Android and Windows.
BitBag
BitBag is Agentic Developer Hub's AI persona for developers. It is a Hub persona: a prompt, a voice and a character, with memory and a knowledge base. It speaks through an animated avatar drawn by a separate engine. It knows the Hub, the Studio's tools and a curated timeline of what matters in agentic development, and it runs on the web and on Apple platforms, Android and Windows, on whichever model and provider it is given.
Agentic Developer Kitchen
Agentic Developer Kitchen is an IDE for working in partnership with AI models. Each project opens in a window of panes the developer arranges (terminals, a file browser, editors, notes, a browser and AI chat) that reopens exactly as it was left, and the app's own daemon and tool server let models drive its browser and terminals while their sessions are recorded and overseen. It runs VS Code extensions and keeps notes and sessions in the cloud through Agentic Developer Hub, with models from many providers on Apple platforms, Android and Windows.
Stenographer
Stenographer is a local recorder for AI coding sessions. One background service captures each session's events and transcript into a SQLite database on the developer's machine, and the app, the command line and the status line are clients of it. It shows live sessions, conversations, cost, charts and oversight rules, tracks working time by project for billing, and runs its AI features on models from many providers. For now it runs on macOS only.
Agentic Developer Community
Agentic Developer Community is Agentic Developer Hub's own discussion forum, a board of topic rooms and threads that anyone can read and Hub members can write in. It holds no data of its own: its topics, markdown posts, nested replies, reactions and roles live in the Hub, which mounts the same forum in every workspace, and agents read it through the Hub's MCP server. A thread stays filed in its room and searchable, members have profiles and a people directory, and moderators and admins pin, lock, mark answers and manage rooms and roles. It runs on the web and on Apple platforms, Android and Windows.
Agentic Developer Docs
Agentic Developer Docs is a private library in Agentic Developer Hub for the documents a team's work produces, such as specs, runbooks, contracts and decisions. A document is the same kind of markdown record as a Hub note or research paper, kept on a shelf of its own. Its pages open in every Hub workspace and on the site. A team files documents in nested categories, tags them across categories, keeps every version, stores files of any type beside them, and searches the full text of everything its people and agents wrote. It runs on the web and on Apple platforms, Android and Windows.
Agentic Developer Research
Agentic Developer Research is a place in Agentic Developer Hub where people and their agents write research documents in markdown and publish them as papers. A paper stays private until its author publishes it, which puts it on a public page under their name that search engines index. The same editor runs in every Hub workspace and on the site. Authors file papers in nested categories and tag them, and agents hand over their findings as ordinary markdown through the Hub's MCP server. Anyone can search the published papers without an account. It runs on the web and on Apple platforms, Android and Windows, and works with agents on models from many providers.