- What does it take to run?
- Self-hosted: a Docker Compose stack with Postgres (pgvector included) on infrastructure you control. Add ollama only if you want local LLM answers, and every feature has a deterministic fallback without it. If you'd rather not operate it, the enterprise plan runs it for you: bring your own cloud, where we deploy and run Mari inside your VPC, or hosted, where we operate the deployment and your team signs in.
- Bring your own cloud or hosted: which should we pick?
- Bring your own cloud if your data has to stay inside your account, or procurement already has a VPC story it wants to reuse: we run Mari there, and you can see and revoke what we reach. Hosted if you want nothing to provision: we run the environment, and where it runs, who can reach it, and how long anything is kept are set out in the MSA and DPA. Same product, same enterprise features, same engineer either way.
- Where does our data go?
- Self-hosted: nowhere. Sources sync into the Postgres you run, and the embeddings and answers come out of that same database. Connectors use tokens you create and can revoke. On the enterprise plan, bring your own cloud keeps all of it in your account, and the hosted option runs in an environment we operate, under the MSA and DPA we sign together.
- Is it open source?
- Yes. Mari is Apache 2.0, and the self-hosted build is free to run yourself: the connectors, cited answers, facts, lineage, drift audits, and the Slack and GitHub bots. The enterprise plan, bring your own cloud or hosted, adds the operations, a dedicated engineer, and features that are not in the open-source build.