The shape of the problem
Most agent platforms ask you to build the agent. Pick the model. Wire the tools. Define the prompts. Tune the loop. Validate the output. Maintain it forever. By the time you have something that works, you have a second job — building the thing that was supposed to do the work for you.
Maestro inverts the trade. We ship the agents. You configure them. We do the work of making them reliable; you do the work of running your business.
One score, three agents, real outcomes
The v1 product is one composed score: B2B SaaS Outbound. Two agents work as a unit:
- Cold leads finds people matching your ICP, drafts personalized openers grounded in real signals (a recent funding round, a hiring post, a migration), and queues them for review or send.
- Reply triage reads inbound mail, classifies intent (interested, not interested, out-of-office, wrong person, unsubscribe, needs-review), advances the contact's stage, and pings you when something needs your attention.
The Pipeline records every action. You see what happened, when, and why — separate from any other system you have. CRM optional. Sync optional. The truth lives here.
Self-hosted, full stop
v1 is a Docker bundle you run on your own box. Postgres, Hono API, Vite web app, Python runtime — all yours. AGPL license. Your inbox tokens never touch our servers because we don't have servers. Your prospect data never leaves your infrastructure.
Cloud SaaS at letmaestro.com is a v2 question. We need to be sure the product works for three design partners before we have an opinion about hosting it for everyone else.
What we will not ship in v1
Scope discipline is what makes a v1 ship. The list of things that are deliberately not in the box:
- External CRM writeback (HubSpot, Salesforce, Pipedrive). The Pipeline is the system of record in Maestro; CSV + webhooks fill the gap.
- Auto-drafted reply generation. Reply triage classifies and notifies. The human writes the actual response. Keeping the human in the loop on every send protects your sender reputation while we build the voice-modeling and feedback loops needed to draft replies that sound like you. Reply drafting is on the roadmap for v2.
- The Composer workspace. v1 is configure-and-run only — operators tune what Maestro ships, they don't build new pieces from the UI. Authoring custom skills, agents, scores, and pipelines is a real product surface that deserves real design; it lands in v1.5+ as a sibling workspace to the catalog.
- Outlook / non-Gmail email. Custom SMTP. Multi-seat / team / roles. Custom fields on contacts. AI-powered ICP builder. Bulk import smart-merge. Public template marketplace. Third-party agent author API.
- The 30-skills-and-community-marketplace platform pivot. Three design partners, one score, real meetings booked. That's the bar.
These aren't bad ideas. They're v2 ideas. Ship the hero score first.
The bar every decision serves
A small business owner signs up on Sunday night, defines their ICP in a textarea, connects Gmail, picks the B2B SaaS Outbound score, and goes to bed. Monday at 09:00 the score runs. By Wednesday they have replies. By Friday they have a booked demo. The Pipeline shows them exactly what Maestro did, separate from any other system they have.
If a v1 PR doesn't serve that scenario, it's v2.
Maestro markets Maestro
The first design partner is the founder. We're using Maestro to find the next two. The cold-leads agent prospects for design partners. The reply-triage agent flags interested replies. The Pipeline UI shows what Maestro brought in for itself. When this works for us, it works as the case study.
Until the founder has personally booked a demo for Maestro using Maestro, we don't open the doors any wider.