What Creux builds, and how we work
Creux designs, builds and deploys AI agents that work inside a business's existing tools. They watch for problems, propose a fix, and act only once a person approves it.
That last part is the point. Most back-office work follows rules that someone on your team applies by hand: does this invoice match the purchase order, is a contract about to auto-renew, is this customer already in the CRM. Software that only waits for instructions leaves that checking to your people. An agent that can act on its own with no review is a different risk. We build the middle path: the agent does the watching and the drafting, and a person makes the call.
Two ways we work
Agentic product engineering. We design and build the agent and the product around it: the tools it can use, the data it reads, the guardrails, the approval flow, and the interface your team works in.
Forward-deployed engineering. Our engineers embed with your team, map how the work really happens, and deploy agents into your real systems. We keep improving them after launch, using what we see in production.
How a project runs
- Map the workflow with the people closest to it.
- Build the smallest useful agent around the clearest constraint.
- Deploy it inside the tools the team already uses, with approval before any action.
- Improve it from production feedback.
We start small on purpose. A narrow agent that people trust and use is worth more than a broad one nobody relies on.
Who we are
Creux is a small team based in Austin, Texas, working remotely with clients elsewhere. Dakota leads the engineering. We take on a limited number of projects so the people designing the system are also the people building and running it.
What to expect from this blog
Notes on building agents that survive contact with real operations: approval design, evaluation, integrations, and what goes wrong. If you have a workflow in mind, get in touch.