Heywire

Heywire started as n8n-with-agents: the same connect-everything automation, but driven by agents instead of rigid schemas. Within days of my finishing routines, the big labs shipped routines and connectors of their own. So I pivoted up, to the thing they aren’t building — one canvas where agents and people work the same surface together. Reading where the frontier is about to land and moving before it does is the instinct this work runs on.

Workspace is the surface — agents and teammates collaborating toward a goal, anything from accounting to content, with an organizational memory (Brain) passed into every task, so within weeks it knows how you work and what your company is. The automation is built from agentic nodes: you wire a flow the way you would in n8n or Weavy, but each node has a little intelligence — you can talk to it — and you keep full control of how work passes step to step. The labs ship conversational agents with no flow control; the automation tools ship rigid schemas with no intelligence; this sits deliberately between. Inside live the Studios — rich-text, content, design, code (desktop, wired to your GitHub), and data — and everything you make is editable by language, by hand, or by an agent, and becomes context for the next thing. And it comes with you: everyone gets a Wire, one personal agent you reach from your phone that docks into the canvas at your desk — so the work leaves the desktop and your context travels with you.

Run a strong model end-to-end through an API and the bill makes automation a non-starter for anyone without an enterprise budget. So Heywire breaks the work down: a capable planner splits a job into subtasks, cheaper models run the parts, a strong model compiles the result — a fraction of the cost for the same delivery. The model is the user’s choice per task, and the system steers them off under-powering a hard job. It’s model-agnostic by construction — and as token economics tighten, that’s exactly what lets a freelancer or an SMB afford AI at all.

It’s an output multiplier for freelancers and SMBs, but the launch wedge is the emerging automation-consultancy market — the agencies and operators the late majority will pay to set their AI up for them. Heywire fits because it’s fast to build in (you direct an agent, you don’t write a code schema) and transparent to hand over: share the canvas as a link and the client can see exactly how their automation is wired, and approve it. The advanced machinery is packaged to feel simple — human-first, even though it’s heavily AI underneath.

Heywire is my idea and my execution, built AI-first: me and the model, together. My hands are on the concept, the architecture, the entire product surface, and the orchestration — I challenge every plan the model proposes before it runs, design the build schemas it works to, and gate everything to a Grade-A bar with confidence and gap checks throughout. I’m not a career Python engineer and I don’t pretend to be; what I have is a real, working understanding of how the modules talk to each other and why — enough to direct the build, keep it honest, and stay in control. My CTO is a call away for the deep questions; my CEO is who I pressure-test the thinking against. And every action still runs inside governance — budget caps, approval gates, an audit trail, a staged path to production — because deploying an agent safely is the actual hard part. Your Wire carries that governance out of the app, too: when a job needs your sign-off, it reaches you wherever you are — not a dashboard nobody opens.

Creator & Builder — the idea, the architecture, and the build, AI-first. A separate venture, same core team as Oyster; my CTO and CEO are my technical and business sounding boards. Honest stage: beta, in investor preview — judged on the hard problems solved, not traction.