Claude Desktop

Connect PleumRouter through the Claude desktop app's third-party inference (gateway) setting.

Claude Desktop is Anthropic's desktop app. In Developer Mode, the Configure Third-Party Inference form takes a gateway address and key, and the app then calls models through PleumRouter. You do not sign in with an Anthropic account this way. For the terminal tool, see Claude Code.

Quick start#

Create an API key
In the dashboard's API keys page, create a key that starts with plm_. A dedicated key just for the desktop app, named clearly, is easy to revoke later.
Turn on Developer Mode
Open the app and choose Help, then Troubleshooting, then Enable Developer Mode. On Mac it is in the menu bar at the top of the screen; on Windows it is in the ☰ menu at the top left of the sign-in screen. Do not sign in with an Anthropic account.

Enter the Gateway settings

In Developer, open Configure Third-Party Inference, set Connection to Gateway, and enter the values below. Do not add /v1 to the address.

Configure Third-Party Inference

Connection            Gateway
Gateway base URL      https://apirouter.pleum.ai
Credential kind       Static API key
Gateway API key       plm_xxxxxxxxxxxxxxxx
Gateway auth scheme   bearer

Apply and restart
Click Apply Changes (Apply locally in some versions) and the app restarts. If a button offers to continue with the gateway (third party), choose it. If there is a model field, enter a catalog ID such as claude-sonnet-5. A reply to a short message means you are connected.

Why use PleumRouter with Claude Desktop?#

Pay only for what you use from prepaid credits. Credits are deducted separately from the tool maker's subscription.
Set budgets and limits per key. Daily, weekly and monthly budgets, model allow-lists and team budgets help you manage what an agent spends.
One key, many models. Pick Claude, GPT, Gemini or open models; automatic retries and fallback cover provider outages; usage shows up in the dashboard.

How it works#

In Gateway mode the app sends Anthropic (Messages) requests straight to the gateway. With a base URL of https://apirouter.pleum.ai it posts to /v1/messages, authenticated by sending your plm_ key as Authorization: Bearer. It is the same endpoint as Claude Code, so PleumRouter behaves the same. The in-app form is the supported way to configure this app; do not guess at files in the hidden Application Support folder.

Configuration reference#

Setting	Notes
ConnectionGateway	Selects the gateway backend.
Gateway base URLhttps://apirouter.pleum.ai	The origin without `/v1`, and no trailing `/`.
Credential kindStatic API key	A fixed API key. Leave the OIDC-related fields blank.
Gateway API keyplm_xxxxxxxxxxxxxxxx	Your PleumRouter API key.
Gateway auth schemebearer	Sent as an `Authorization: Bearer` header.
Allowed egress hosts(if needed)	Add hosts here if web fetching (WebFetch) is blocked.

Choosing models#

If there is a model field, enter a PleumRouter catalog ID as is. Look up exact IDs in the model catalog or with GET /v1/models.

These steps assume Claude models (claude-sonnet-5, claude-opus-5-5 and so on). Other makers' models connect too, but some features can behave differently.

Compatibility#

Feature	Status	Notes
Streaming and tool calling	Yes	Uses the same `/v1/messages` endpoint as Claude Code.
Image input	Yes	Works only when the selected model accepts images.
SSH, Remote Control	No	Not offered in gateway mode. That is a limit, not a fault.
Web fetching (WebFetch)	Partial	The sandbox may block it, so add the host under Allowed egress hosts.
Features tied to a Claude account	Unverified	You do not sign in with an account, so account-linked features may not be available.
Non-Claude models	Partial	They connect, but it is not a combination Anthropic supports.

Troubleshooting#

Symptom	Cause and fix
The Gateway option is missing	After turning on Developer Mode, apply the settings, fully quit the app (Cmd+Q on Mac) and reopen it.
Connection errors	Check that Gateway base URL is exactly `https://apirouter.pleum.ai`. A `/v1` or a trailing `/` makes it fail.
401 Unauthorized	Check that Credential kind is Static API key, the auth scheme is bearer and the key starts with `plm_`, and leave the OIDC fields blank.
No models, or model not found	Confirm the catalog ID with `GET /v1/models`. If a 402 appears too, check your credit balance and the key's budget.
Web fetching is blocked	In Developer's Configure Third-Party Inference, add the host to the Allowed egress hosts field.
A sign-in screen appears after reopening	Choose the button that continues with the gateway instead of signing in to Anthropic. If the settings vanished, apply step 3 again.

Cost & security tips#

Set a budget on the key. A coding agent calls the model many times per task and resends long context. Daily, weekly or monthly budgets on your API key make costs easier to predict.
Keep the key out of git. Make sure a key in your shell profile or config file never gets committed, and if you suspect a leak, revoke it in the dashboard and issue a new one.
Check where sensitive code goes. Prompts and code are sent to the provider of the model you pick. Use Data residency and PII masking to control that.