The local endpoint

While Your Own AI is open, it serves an OpenAI-compatible endpoint on your machine:

http://localhost:11435/v1

Any app that can talk to an OpenAI-style API - agent frameworks, editors, a ten-line script - can use your AIs: the personality you authored answers, your memories inform it, and the conversation lands in your signed records like any other, marked with an "API" badge on the Memory page.

Setup in the app

Settings → External access shows the endpoint, its status, and a Copy setup button per AI - base URL, model name, and a runnable example. No API key is needed on your own computer: being on your machine is the credential.

The model names

Every AI has a model name (its name, slugified). Two variants matter:

Model	What answers
<ai-name>	The AI in full - persona, memory, records
<ai-name>:agent	The same AI tuned for tool-driving agents - plain output, no persona flourish

A minimal call

curl http://localhost:11435/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "maeve",
    "messages": [{"role": "user", "content": "Tell me a short story."}]
  }'

Streaming, multi-turn conversations, and tool calls follow the standard OpenAI shapes. See Connecting apps and agents for framework recipes and the request headers reference.

Other devices on your network

By default the endpoint is loopback-only - reachable from this computer, invisible to the network. To serve your AIs to your other machines, turn on Settings → External access → Other devices on this network. The app mints an access key and shows the network address; on the other device, use that address as the base URL and paste the key as the API key - the same three fields as ever. Devices without the key are refused, local apps stay keyless, and you can mint a new key anytime.

What runs where

Calls route like any other question: your AI's mode decides (routing applies), so an offline-only AI answers offline-only - external apps included.