Useful for remote model access, but they are not the same thing as a self-hosted operator runtime.
Self-hosted local AI runtime
You host everything. Heiwa ships software, not a hosted operator app.
Reuse the AI subscriptions you already pay for.
Heiwa turns Claude Max, ChatGPT Plus, Google AI Pro, API-key providers, and local Ollama into one operator runtime you host yourself, with routing, receipts, approvals, and evidence.
Canonical install path
curl -fsSL https://heiwa.ltd/install | sh
Why Heiwa exists
Not another hosted router.
Heiwa is for operators who want one runtime across local models, provider CLIs, and API-key providers without handing the runtime, app, memory, or secrets to someone else's hosted surface.
The website delivers software. The operator machine runs the runtime, app, REPL, and secrets.
Three provider lanes
Ollama and machine-owned models
Private, fast, and cheap by default. Local inference stays on the operator's machine.
Claude Max, ChatGPT Plus, Google AI Pro
Reuse subscriptions you already pay for through provider-owned CLIs under one routing surface.
OpenAI, Anthropic, Google, OpenRouter, and more
Bring keyed providers into the same runtime when remote APIs are the right escalation path.
Operator flow
Start from a one-line install or GitHub Release. The runtime lands on the operator machine.
Authenticate provider CLIs, add keys, and discover local models. Heiwa normalizes the lanes.
Use the REPL, open the local app with heiwa app, and keep evidence on the machine.
Shop window only
heiwa.ltd delivers software. Your machine runs Heiwa.
The public site is for install, docs, release delivery, and identity exchange. Operator work, memory, sessions, and provider secrets stay with the runtime you host yourself.