About Olivares AI
Olivares AI is an open, self-hostable platform to run, integrate and govern the AI that works on the infrastructure you operate — Linux, Docker, on-prem, multi-cloud, air-gapped — with Claude Code, Codex, Grok Build and the models you host yourself. Multi-provider by design, from a home server to a regulated enterprise.
Why we exist
AI stopped being one chat window. What an organization runs now is coding agents in terminals, MCP servers, model endpoints and service accounts, spread across machines nobody designed as one system. The teams who feel that first are the ones who run the infrastructure those agents touch, and we build for them: platform, DevOps and security teams — and the advanced users running the same thing at home. The product gives them one durable plane for the work itself (who holds a task, what it may reach, what it actually reached) and the evidence to prove it afterwards.
Independent research is blunt about the gap:
What we won’t compromise
A platform that sees this much of your estate has to be secure and auditable itself — your failure would become the customer’s breach. Security comes first, not as a later hardening pass, and the documentation says plainly what runs today, what is design-stage and what the product deliberately does not do.
Claude Code, Codex, Grok Build, gemini-cli, Cursor, opencode, goose, cline, OpenHands, OpenClaw and Hermes — and the models you host yourself. Each surface states what it can enforce and what it can only observe; none of them is the product’s centre of gravity, and each one is integrated as deep as its tools allow.
The AGPL build is the whole platform — never feature-capped, unlimited accounts. The commercial add-ons are additive code, and a subscription is the credential you download signed artifacts with, the way the enterprise Linux distributions settled it. The community edition is what most people will run, and that is the point.
One binary, SQLite or Postgres, on Linux, Docker, Kubernetes or fully air-gapped; policy, budgets and access decided per agent, session, group and surface. The project is at home in the Linux and open-source community, and it is built to run where that community runs.
Who it’s for
It is useful to one person on day one: an advanced user or a sysadmin who wants to see what is running and hand agents real work without losing track of it. And it scales without changing shape — a freelancer with a tenant per client, an engineering team sharing work items and leases, a regulated enterprise on Postgres with row-level security and air-gapped installs. The same binary, with nothing withheld.
Olivares AI was founded by Fran Olivares — Francisco Olivares in full — and is built in Spain, almost entirely with AI: parallel coding-agent sessions in isolated git worktrees, coordinated through the repository itself — the problem the product now solves. The project is deliberately on the side of AI — integrating it, distributing it and making it usable in every kind of environment, for every kind of person, team and company — and it is at home in the Linux and open-source community. The founder statement and press materials live on the press page.
Founder and press materialsDeploy Olivares AI on your own infrastructure, or read the code to see exactly how it works.