Your agents. On your terms.
Fast. Less baggage. No BS.
Build an agent stack that works for you. Give your agents a home, bring them into your chat, and keep the pieces you want.
Explore the Soulsteads lineup →
- Agent Home — A portable home for your agents, with identity and history.
- Agent Messaging — Bring coding agents into your chat. Available as an early release.
- Agent Forge — Part of the upcoming lineup. Preview only; no public implementation yet.
- Model Router — A Rust gateway for routing OpenAI-compatible requests across your own inference servers, with stable aliases, capability checks, and shared capacity limits. Early release; builds as
slotline.
Visit the Soulsteads website to explore the products, see what's available, and get started.
Model Router gives compatible clients one endpoint for existing inference servers. It does not host models or manage loading, and it is not a universal OpenAI drop-in. Use the setup guide and compatibility guide before connecting clients.
You need a Linux/Unix host, an existing inference server, and an assigned Tailnet
IPv4 address. The listener rejects loopback, LAN, and wildcard binds and reserves
port 4001. It serves HTTP without built-in TLS; keep it on a protected network,
not exposed directly to the internet. No model server or prebuilt binary is bundled.
The project's release notes still require a Git-history privacy review before
public-release qualification; published source is not a claim that every release
gate or backend combination is qualified.
Home, Messaging, and Model Router are independent tools. Installing one does not configure another, migrate clients, or start a service.