Every feature with its state and owner — and, where the team asked for it, the review it is waiting on.
The hero automation. It finds people matching your ideal customer, drafts a personal opener for each, sends from your own inbox, reads and classifies the replies, and records every contact and every interaction in a pipeline you own.
Two graphs: one for outbound, one for reply triage. Both retargetable by editing plain-English instructions — no code change needed to point at a different kind of customer.
Nothing reaches a customer without a person approving it. Drafts queue up as cards, you read them, and only then do they send.
On by default. It is the reason an automation can be trusted with a real inbox.
A canvas for building automations. Drag steps on, draw the connections between them, configure each one in place, and save.
Replaced the configure-only era, where changing an automation meant editing seed data.
Attach an automation to a schedule and it runs on its own. Every run is recorded step by step, so you can see exactly what happened and what it cost.
Contacts and their history live inside Maestro, so there is no CRM to buy before the first automation can run. Everything is exportable.
Deliberately separate from whatever CRM the business already has. Maestro owns its own record of what it did.
Reusable agent setups — the instructions, the model, and the tools it is allowed to touch — that any automation can point at.
Seven ship in the box: lead search, email, web research, writing, notifications, the pipeline itself, and a general escape hatch for anything else with an API.
Each one declares what it can do and what credentials it needs, and the runtime finds them at startup.
The whole thing runs on one machine you control, behind your own domain, with no ports open to the internet. Sign-in happens at the edge.
Docker Compose plus a Cloudflare tunnel. Migrations and first-run setup happen on their own.
A written contract for how automations, agents and integrations get authored, in a form both a person and a machine can read — so everything built later is built the same way.
Say what you want the automation to do and have it drafted for you. The first of four steps is done: the authoring contract, and a single checkable definition of what a valid automation is.
The next step is an in-browser editor for integrations, bound to the published schema.