How-To · 10 min

Lend models between your machines

Run Kaptain on two machines and each can use the other's models — a beefy desktop serves its local models to a laptop that could never load them. One code grants it; deleting the code ends it.

What crosses (and what never does)

Lending shares the owner's models: the borrowing machine can run inference on them, can send an image across for the owner's model to describe (the picture crosses the link once), and follows the model settings the owner set for them, such as a context window. It cannot see the owner's chats, files, or anything else. A machine still on 0.0.7 lends its models but is shown disabled for image processing with "update it". Transport is your tailnet — both machines must be on it (Tailscale guide).

1 · Issue a connection code

1. On the machine that owns the models: Settings ▸ Network Access ▸ Share models between your computers ▸ Issue a token.
2. Create a connection code — it looks like kcat_…. Copy it now; treat it like a key.
3. Note the owner machine's tailnet address (shown in Network Access).

2 · Connect from the borrowing machine

1. On the other Kaptain: Settings ▸ Network Access ▸ Share models between your computers ▸ connect to another Kaptain.
2. Paste the owner's address and the kcat_… code.

3 · Verify

Open the model picker on the borrowing machine: the owner's models appear grouped under that computer's name, usable everywhere a local model is — chat, kapsules, delegation. Run one small prompt on a borrowed model to see it answer.

4 · Revoke

Delete the code on the owner machine (Share models between your computers) and the borrower loses it immediately. Nothing to clean up on the other side.

kcat_ vs chynj_ — don't mix them up

Agent-readable version: index.md · Every claim on this page was exercised on a real install before being written. Something missing that cost you time? That is a documentation bug — tell us.