Platform
Run each agent on the model that suits it, from the frontier to open models on your own GPU, with the same memory and tools on every one. Then let them work around the clock, connected to the systems you choose, on every screen you own.
Pick the model per agent, or per task. Nothing else changes.
Use a Claude Pro or Max subscription, a ChatGPT subscription through Codex, or API keys for Anthropic, OpenAI, Google Gemini and Perplexity. The model picker marks which models bill your subscription and which bill per token.
A built-in proxy runs open models with llama.cpp or NInfer, loads each one on demand and swaps them on a single GPU. LM Studio and Ollama work too. Local turns are free and never leave the house.
Memory, records, recall, consultations, jobs, schedules, files and the permission gate work the same on every model. Switching never costs a capability.
Change an agent's model, or how hard it thinks, from the chat header on desktop or phone. When a subscription nears its limit, the hub steps down to the next model and tells you.
Your agents keep working when you stop typing.
In plain language, once or on repeat: tomorrow at nine, every weekday at 8:15, every Monday. When the time comes, the agent wakes with the task and reports back.
Long work such as a build, a data pull or a backtest runs on its own. When it finishes, the agent is woken with the result and carries on from there.
Mid-task decisions arrive with buttons and a text box, and they wait for you through a closed laptop, a lost signal or a reloaded page.
See which agents are busy, which tools they are using and what each worker is doing, live. Add a note mid-task and the agent takes it on board, or stop it from any device.
Agents reach the systems you choose, with credentials the hub holds for them.
Plug in any MCP server: a mailbox, a document store, an internal API. Credentials are encrypted, can be locked behind a passphrase, and are granted to named agents only.
Search through Brave, Perplexity, Serper, Tavily, your own SearXNG, or DuckDuckGo with no key at all. Pages come back as clean text, PDFs are read, and pages that need JavaScript are rendered first.
Agents can message you on Telegram, write from an email address of their own, and work with Google Workspace, GitHub and more.
Give a script a token scoped to exactly what it needs, such as waking one agent when an alert fires. Other applications can talk to your agents through an OpenAI-compatible endpoint.
One hub, at your desk or in your pocket.
A full desktop app and an installable phone app with the same agents, history, files and controls. Push notifications tell you when an agent finishes or needs you.
Agents read photos, screenshots and scanned PDFs, draw charts and diagrams right in the conversation, and create illustrations with image models.
Every turn is costed. Subscription and pay-per-token usage are labelled apart, and subscription limits are shown as you use them.
Update from the toolbar or the phone menu, choosing the stable or the experimental channel. The hub installs, restarts and confirms it came back.
How My Maestro AI keeps capable agents safe, private and under your control.
My Maestro AI