PandaStack
Self-hosting

Self-host overview

Run PandaStack on your own infrastructure — GCP, AWS, on-prem bare metal, or locally on Linux and Apple Silicon.

PandaStack is open source under the Apache-2.0 license, and everything the hosted platform runs — the control-plane API, the scheduler, the per-host agents, and the Firecracker sandbox hosts — can run on infrastructure you own. Common reasons to self-host:

  • Open source, no lock-in — the full stack is Apache-2.0. You run the same code as the hosted platform and can audit, patch, and fork it.
  • Data residency and compliance — sandbox workloads, filesystems, and snapshots never leave your environment, which matters when code or data can't touch third-party infrastructure.
  • Run in your own VPC — deploy the API and agents inside your private network, next to your internal services and data stores, with your own network policy and no public egress requirements.

Architecturally, a self-hosted deployment is a control plane (API + Postgres), one or more agents on KVM-capable Linux hosts, and the Firecracker microVMs they manage. You run those services and point clients at your API endpoint.

Deployment guides

Pick the guide that matches your target:

  • Deploy on GCP — Terraform on Compute Engine, using nested virtualization on standard machine types; single dev host or multi-node topology.
  • Deploy on AWS — bare-metal EC2 (c5n.metal) via the dev Terraform environment and the make tf-* targets.
  • Deploy on-prem — run the API, dashboard, scheduler, and agents on your own bare-metal hardware.
  • Deploy on Mac (M1/M2/M3) — quick Apple Silicon setup via Lima, for hacking on PandaStack itself.
  • Local install on Linux — end-to-end on a standalone Ubuntu/Debian box with one command; Firecracker runs directly on the host.
  • Local Apple Silicon development — the full Firecracker stack on an Apple Silicon Mac with Lima and nested virtualization, in depth.
  • Supabase auth — point a self-hosted API at your Supabase project's JWKS so dashboard sign-in produces verifiable JWTs.

Client configuration

Use the same SDKs and CLI against your endpoint:

export PANDASTACK_API_KEY=pds_abc123def456...
export PANDASTACK_API=https://api.pandastack.ai

For a private deployment, replace PANDASTACK_API with your HTTPS API origin.

Install the CLI

npm install -g @pandastack/sdk   # provides the `pandastack` command

Smoke test

pandastack me
pandastack template list
pandastack sandbox create --template base --ttl 3600
pandastack sandbox list

The CLI does not deploy the platform itself; cluster deployment automation lives outside the pandastack command surface.

On this page