Portfolio · Built In-House

The products we build & run

We hold ourselves to the standard we sell. These two flagship systems run our own operation every day.

A terminal-native operating system for the business, and a fleet of seven autonomous AI agents that work the sales funnel around the clock. Designed, engineered, and operated by the same team you’d be hiring.

Product 01

Commandeer OS

A local-first, terminal-native operating system for running an independent consultancy — no SaaS subscriptions, no cloud lock-in, no monthly creep.

Everything a services business needs to generate documents, track clients, and move fast lives behind one elegant, colour-coded CLI. It’s composable by design: a module registry lets us bolt on new capabilities without touching the core.

python3 commandeer.py
╔════════════════════════════════════════╗
║   ⚓  COMMANDEER CONSULTANCY  ⚓        ║
╚════════════════════════════════════════╝

  AVAILABLE MODULES

  1 ▶ Invoice Generator
  2 ▶ Proposal Generator
  3 ▶ NDA Generator
  4 ▶ Client Manager
  5 ▶ Profile Settings
  0 → Exit Commandeer

  Choose module (0-5): 

Document generation

Invoices, proposals, NDAs, SOWs and quotes as PDF + DOCX, rendered from Jinja2 templates and auto-numbered.

Client manager

SQLite-backed client records — company, contact, day rate, status, notes — with full CRUD from the CLI.

Voice-note drafting

Record a voice note, transcribe with Whisper, pipe the transcript to Claude, and get a structured proposal in your template.

Composable modules

A registry-driven module engine. Every tool is self-contained and lazy-loaded — pick your next build from the catalogue.

Built withPythonSQLiteJinja2python-docxfpdf2WhisperClaude
Product 02 · Operation Autonomous

Commandeer Fleet

A tenant-agnostic fleet of seven autonomous agents that run the top of the sales funnel 24/7 — for any number of clients you configure.

They prospect, research, write, send, and triage cold outreach without supervision. There are no hardcoded businesses: each client is a tenant with its own ICP, data sources, sending limits, and voice. A human only enters the loop when a prospect replies interested.

The lead lifecycle
NEW
RESEARCHED
SEQUENCED
REPLIED
HOT 🔥
07:00 UTC · daily

Prospector

Pulls raw leads from every enabled source (Apollo · Apify Maps · RapidAPI), then dedups, verifies, scores against the ICP, and inserts qualified leads as NEW.

every 5 min

Researcher

Turns NEW leads into RESEARCHED ones, gathering context from multiple sources (each fails soft) and distilling a personalised hook.

every 5 min

Writer

Turns RESEARCHED leads into a written, persisted outreach sequence — composed by Claude and matched to each client’s voice samples.

every 30 min

Sender

Guards deliverability for every sending campaign and automatically pauses a client the moment bounce or complaint thresholds are breached.

every 2 min

Classifier

Polls replies, classifies intent, and routes. INTERESTED / REFERRAL becomes a HOT lead and fires an instant Telegram alert.

Mon 07:00 UTC

Analyst

Aggregates each client’s real metrics into a weekly report — numbers computed directly, prose written by Claude with a templated fallback.

Mon/Wed/Fri 08:30 UTC

Content

Builds the inbound engine: drafts a LinkedIn post per client from a real news trigger and sends it for human approval.

Roadmap

Voice & Phone Scripting

An eighth channel in progress: AI voice agents that run scripted outbound calls, qualify on the line, and hand warm prospects straight to a human — joining email and LinkedIn in the sequence.

Deliverability-safe

The Sender watches bounce and complaint rates per campaign and auto-pauses before reputation takes a hit.

Human-in-the-loop

Interested replies and referrals fire an instant Telegram alert — you act only on warm, qualified intent.

Multi-tenant by design

A single worker owns the Postgres database and runs every active client each cycle. Add tenants from the dashboard.

Fleet Dashboard

A thin Next.js control panel — the fleet’s bridge. Configure every tenant (ICP, data sources, sequence, sending limits, notifications), then monitor pipeline, agent health, hot leads and deliverability live. It holds no database and talks to the worker over a server-only REST API, so the key never reaches the browser.

Built withTypeScriptFastifyPrismaPostgreSQLNext.jsClaudeSmartleadTelegramZodpm2
Why this matters to you

Proof, not promises

We architect multi-tenant systems, autonomous agents, and AI pipelines — for ourselves first, then for you.

Every pattern here (typed config, fail-soft integrations, server-only API boundaries) is the same discipline we bring to your codebase.

When we recommend an approach, we’ve already run it in production against our own revenue.

Want this kind of system in your business?

Whether it’s an internal operating system or an autonomous agent fleet of your own, we build it end to end. 30 min call, no pitch.

Ready to set sail?

Let's chart your course to success