Skip to content

What are self-hosted AI agents?

Self-hosted AI agents run on infrastructure you control, such as your own servers or private cloud, instead of a vendor's multi-tenant cloud. The data, the agent runtime, and optionally the language model all stay inside your boundary.

This matters most for regulated industries, and for any organization that cannot send its prompts and documents to an external service.

| | Cloud AI agents | Self-hosted AI agents | | -------------------------- | ------------------------------------ | --------------------------- | | Operated by | The vendor, on shared infrastructure | You, on your infrastructure | | Where data goes | To the vendor's servers | Stays in your boundary | | Time to first run | Lowest | Higher (you deploy it) | | Data residency / isolation | Vendor-controlled | You control it | | Offline / air-gapped | Not possible | Possible with local models |

Cloud is the easiest way to start. Self-hosting is what you choose when "where does the data go?" has a mandatory answer.

For teams under GDPR, sector regulation, or the EU Cloud Act, sending prompts and documents to a US-operated cloud is often a non-starter. Self-hosting removes the question entirely: the data never leaves. That is an architecture decision, not a settings toggle, which is why it has to be designed in from the start rather than bolted on.

Paired with a local model runtime such as Ollama, a self-hosted agent platform can run with no external network calls at all, which is what air-gapped and high-compliance environments require. When offline operation is not a requirement, the same platform can use cloud model APIs.

Pinchy is self-hosted first: a single Docker Compose stack, pre-built images, and support for local models via Ollama for fully offline operation. The license key is validated offline, so the instance never phones home. See Self-hosting on the main site and the Installation Guide.