About us
Model access should be a single dependency.
vibe-txt exists because the AI ecosystem fragmented faster than any team could keep up with. Six providers, six SDKs, six billing portals, six sets of rate limits — and a new frontier model every other week. We collapsed all of it into one key, one endpoint, and one command.
Our story
It started as an internal script
In the beginning of 2026 our founding team was shipping an agent product and burning most of its engineering time on plumbing: normalizing streaming formats, retrying across providers during outages, and reconciling five invoices at the end of the month.
The fix was a 400-line proxy that gave every service one endpoint and one key. It spread across the company in a week, then to friends at other companies, then to a waitlist we never advertised.
Today that script is a production gateway routing billions of tokens a month for teams that would rather build products than maintain adapters.
Values
What we optimize for
Transparency
Per-token prices, routing decisions, and margins are visible in the dashboard. No hidden multipliers on upstream cost.
Developer-first
Everything starts in the terminal. If a feature can't be driven from the CLI, it isn't finished.
Reliability
Automatic failover across providers and regional redundancy. If a model or region is down, requests keep moving.
Openness
OpenAI-compatible endpoints, open-source SDKs, and no proprietary lock-in. Leaving should be as easy as arriving.
We're hiring builders
Remote-first across Europe. If routing traffic at low latency and obsessing over developer ergonomics sounds fun, talk to us.