Just finished a migration project where the team needed Continue but couldn't use the default Claude setup. Got it working with a private, self-hosted Anthropic endpoint instead—took a bit of tinkering.
Here's what worked for me. First, you'll need to add a new model in Continue's `config.json`. The key is pointing `apiBase` to your internal URL and keeping the `apiKey` field (even if your endpoint doesn't require one—just put a placeholder). Then set `model` to whatever your endpoint expects, like "claude-3-5-sonnet". The `provider` should stay "anthropic". Had to restart VS Code a couple times for changes to stick.
Biggest hiccup was the API format—our private endpoint had slight differences from Anthropic's official one. Wrote a tiny middleware to translate, but YMMV. Once it was up, the ROI was immediate: no data leaving our walls and much faster context loading. Team's feedback was super positive after a short training call. Let me know if you hit similar snags!
—j
Trust the trial period.
That "immediate ROI" claim is doing a lot of work. Did you factor in the dev hours for the middleware, the ongoing maintenance for that private endpoint, and the training call you mentioned? What was the actual spend on the private endpoint versus the standard API?
Often the "no data leaving our walls" argument gets used to justify costs that would never pass a procurement review if you just did the math on engineer time. The faster context loading is nice, but you're trading a predictable SaaS spend for a hidden internal capex.
Show me the unit economics.