Monocursive · AI systems engineering

I build complex AI systems that keep working in production.

Agents, MCP servers, durable workflows, product integrations, and deployments—from architecture to observable operation.

Building Hubfluencer , Benzaiflow , and Ouroboros .
Shipping software since 2012.

Discuss your system Explore the systems
Operational AI system architecture showing clients, gateway, agent layer, integrations, data, observability, and production signals.
Architecture pattern Designed for operation, not just launch

Proof

Products and runtimes.

Live products and open-source research, with the engineering behind the interfaces: agents, workflows, and durable state.

Ouroboros

Experimental · Open source

Keep the work alive.

An experimental, distributed runtime for AI coding agents. Recover sessions, verify recorded runs, and place work across your own machines—with your choice of model and a shared view of the work.

  • Durable sessions
  • Verified replay
  • Distributed agents
  • Signed WASM
Platforms
macOS · GNU/Linux
Interfaces
Terminal · Web UI
Runtime
BEAM clustering · per-agent Git worktrees
Explore Ouroboros
Agent runtime Development capture
Ouroboros terminal showing a recovered coding-agent conversation, session history, and two connected fleet nodes.
Real recovery session · two local nodes on one Mac

Your machines. Your models. A session that can continue.

Hubfluencer Studio

In production

Real app recordings. Launch-ready ads.

A native Mac app that turns recordings of your product into finished ads. Capture the real app, let your coding agent shape the edit, and export locally—with cuts, zooms, captions, and a timeline you control.

  • Real app capture
  • Agent editing
  • ACP · MCP
  • Local export
Platform
Native macOS 15+
Editing
Claude Code · Codex · undoable revisions
Workflow
Capture → edit → export on your Mac
Explore Hubfluencer Studio
Native editing studio macOS
Hubfluencer Studio on macOS, showing a real app recording in a vertical ad, an editable timeline, and appearance controls.
Real Studio capture · editable timeline and motion controls
  1. 01 / CaptureDisplay, window, region, or iOS Simulator.
  2. 02 / EditYour agent proposes a cut. You review and refine.
  3. 03 / ExportRender on your Mac. Ready for your launch.

Benzaiflow

In production

Your day, without the dread.

A gentle AI companion that turns a wall of to-dos into a few calm, doable things—then keeps the next move close through chat, calendar, widgets, notifications, and live focus.

  • Per-user agent
  • Durable state
  • Calendar
  • Ambient UI
Surfaces
iOS · Android · Web
Runtime
Agents · Durable Objects
Workflows
Tasks · planning · memory
Explore Benzaiflow
Ambient system Live product

One calm next move, wherever you are.

Benzaiflow notification showing one next move.
Benzaiflow focus widget with pause and done actions. Benzaiflow live activity for an active focus session.

Agent context → useful action → ambient surface

Expertise

From agent behavior to production operation.

One senior builder across the seams where complex AI projects usually break.

01

Agent and workflow systems

Agent runtimes, durable workflows, tool use, review gates, recovery, and human checkpoints.

Agents SDK · state · queues · evaluation · guardrails

02

MCP and product integrations

MCP servers and agent kits that expose real product capabilities through stable, inspectable contracts.

MCP · auth · schemas · APIs · webhooks

03

AI product engineering

Production interfaces and backends where AI is part of the product—not a detached demo.

Web · mobile · realtime · media · data

04

Production hardening and deployment

Idempotency, observability, billing invariants, security, migrations, and releases that survive reality.

Cloudflare · containers · Postgres · CI/CD · tracing

Operating model

Finite scope. Visible state. Honest proof.

I work best where architecture, implementation, product judgment, and release responsibility need to stay connected.

See engagement formats
  1. 01

    Find the real constraint

    Map the product, data, model, infrastructure, and human decisions before choosing an architecture.

  2. 02

    Build the smallest complete system

    Ship a vertical slice with real state, failure paths, review points, and operational ownership.

  3. 03

    Prove it under pressure

    Exercise retries, race conditions, stale state, billing, recovery, and deployment before calling it done.

Depth

Shipping software since 2012.

AI products today, with a decade-plus foundation across energy, e-commerce, fintech, IoT, media, security, and product infrastructure.

  • EDF
  • PrestaShop
  • Lunii
  • Innovorder
  • Combo
  • Mylight Systems

Writing

Notes from the systems.

All writing
01

Introducing Ouroboros: a runtime for AI coding agents

 02

Durable agent execution is a state and authority problem

 03

How to build workflows for AI video ad generation

Have a difficult system?

Let’s make it operable.

Bring the architecture, the failure mode, or the ambitious product. We’ll find the smallest responsible path to production.

Discuss your system contact@monocursive.com