DurinDoor
Integrations

Integrations

Point coding tools at http://localhost:20128/v1 with a DurinDoor key and a model name.

Connect a coding tool after DurinDoor is running, a provider is connected, and a first request succeeds. You need a DurinDoor key and a model ID, alias, or combo that this instance serves.

Choose the client URL

ClientBase URL for a local gateway
OpenAI SDKs and most compatible clientshttp://localhost:20128/v1
Anthropic SDKshttp://localhost:20128
Claude Codehttp://localhost:20128/v1
Cline configuration written by DurinDoorhttp://localhost:20128
Roo with Ollama providerhttp://localhost:20128/v1
CursorA public gateway URL ending in /v1

Clients differ in whether they append /v1 themselves. Follow the tool's page. For a development server, replace port 20128 with 20127. A remote client needs an address it can reach.

Configure the tool

Open CLI Tools and choose the tool card. Apply writes supported tool settings on the host running DurinDoor. If the editor runs on another machine, use Manual Config there. Cursor, Roo, and Continue are guide cards and do not write settings.

Select a real DurinDoor key. Local cards may offer sk_durindoor when no key is selected; this placeholder fails when API-key enforcement is enabled.

Auto-configure is a separate page for local Headroom, PXPIPE, and Firecrawl services. It does not configure coding-tool files.

Confirm traffic

Send a short prompt from the tool and find its request on Usage. A model list alone does not verify inference. Then test the tool calls, context size, and streaming that your workflow needs, especially with fallback combos.

For authentication errors, use the gateway key rather than the upstream vendor key. For missing requests, check the active provider selection and the tool's URL. Codex and some older configurations still use the provider slot 9router; keep that slot where the tool guide requires it.

On this page

Edit on GitHub