Why this exists
Channel teams keep trying to put one model in the middle of a network. The center cannot see a partner’s books. The partner cannot operate without a wall around their data. A generic chat tool does not fix that. Hiring a project team for every partner does not scale either.
OrbySoft sits in between. Each partner gets an agent, memory, a sandbox, and a cap. You run the brand. We run the machines.
What we actually build
The product is the isolation layer — not a slide about “AI transformation.” Agents, jobs, and sandboxes. If we cannot keep two partners apart, we do not have a product.
- Agents talk to Gemini through Vertex AI, unless you point us at another model.
- Functions and sandboxes run on Cloud Run or GKE.
- Usage and evals land in BigQuery.
- Secrets go through IAM and Secret Manager.
- Default region: São Paulo.
How we get paid
A subscription for the platform, then usage for tokens and compute. One invoice for the operator, with a line per partner if you want it.
We are early. The first networks are the ones that tell us what to build next.
Who builds it
Josué Moura Santos
Josué started OrbySoft after watching channel teams share one bot across dozens of partners. Someone always saw the wrong context. The product is the wall.