Create two PATs
Keep publisher and consumer credentials separate. Revoke either one instantly from your dashboard.
Secure outbound tunnels. Zero inbound ports.
Give Codex, your apps, and your team secure access to LM Studio and private services running on your own hardware—without opening your firewall.
LM Studio or a private API listens on your machine. The connector publishes only the endpoints you allow.
local endpoint → 127.0.0.1:1234Three steps
The connector makes one authenticated outbound WebSocket connection. SocketSpark handles the hard parts in between.
Keep publisher and consumer credentials separate. Revoke either one instantly from your dashboard.
Publish LM Studio or any allowlisted TCP endpoint. Nothing listens on your public network.
Point OpenAI-compatible tools at SocketSpark Connect and authenticate with the consumer PAT.
Built for work in motion
Use the GPU workstation in your office from a laptop at a café. Connect an automation runner to a private development API. Reach a database without exposing it to the internet.
curl 'https://connect.socketspark.com/v1/models' \
--header 'Authorization: Bearer REPLACE_WITH_CONSUMER_PAT'
Requires an online LM Studio publisher and a separate consumer PAT.
01 / Network boundary
The connector initiates the encrypted connection. Requests return through that established channel; your router never needs an inbound port-forwarding rule.
02 / Credential boundary
Publisher and consumer PATs must be different. Keep each credential private, set an expiry, and revoke access from your dashboard.
Ready when you are