Back to guides

Anvil Integration

Connect Anvil to LLM Gateway with one-click browser login — DevPass or pay-as-you-go — and run repo-aware agent sessions with any model.

Anvil is a chat-first desktop workspace for repo-aware agent delivery. It keeps conversations, repositories, work items, Git state, reviews, and terminals together while agents do the work. LLM Gateway is a built-in provider with one-click browser login — no keys to copy.

Two Billing Modes

Anvil's LLM Gateway connector has two billing modes sharing one login flow. DevPass lists canonical models (claude-opus-5) that the gateway routes for you on a flat coding plan; Pay as you go lists provider-pinned models (anthropic/claude-opus-5) billed against your credits. Pick the mode matching your account before connecting.

Using DevPass? Keep the default DevPass mode. Provider-pinned routing is not available on coding plans, so canonical model IDs are the ones that work.

Prerequisites

Setup

1. Open the LLMGateway Connector

Launch Anvil and open Settings. In the provider list, select LLMGateway ("DevPass or pay-as-you-go models through one gateway connection"). The same connector is also offered during first-run onboarding.

2. Pick a Billing Mode

Choose DevPass for a coding plan or Pay as you go for credits. The mode decides which model catalog Anvil loads.

3. Connect in Browser

Click Connect in browser. Anvil opens the LLM Gateway authorization page — check the account and organization, then approve the connection. Approval creates an API key and returns it to Anvil through a local callback; keep Anvil running until the callback finishes (the login times out after 5 minutes).

Prefer a key you already have? Paste it into the API key (alternative) field instead — keys start with llmgtwy_.

4. Select a Model

Pick a model for the session. Anvil lists the gateway's agent-capable (tool-calling) models for the selected billing mode, with per-turn reasoning-effort controls where the model supports them. Check the live catalogue for current capabilities.

5. Start Building

Run planning, implementation, or review sessions grounded in your checked-out repositories. All requests route through LLM Gateway and show up as Anvil in your dashboard's agents view, with usage, costs, and logs.

Why Use LLM Gateway with Anvil

  • One login for every model — Claude, GPT, Gemini, DeepSeek, and more through a single connection
  • DevPass or credits — flat-price coding plan or pay-as-you-go, switchable in Settings
  • Cost tracking — see exactly what each agent session costs in your dashboard
  • Automatic fallback — if a provider is down, requests route to an alternative
  • Volume discounts — check discounted models for savings

Troubleshooting

Login times out

The browser login expires after 5 minutes. Click Connect in browser again to get a fresh authorization link, and complete the approval while Anvil is running.

No models listed

Anvil only lists agent-capable (tool-calling) models for the selected billing mode. Switch the billing mode to match your account type, or refresh the connector status from Settings.

Authentication errors

Make sure the connected key is active — manage your organization's API keys in the dashboard. Disconnect and reconnect from Settings to mint a fresh key.

Need help? Join our Discord community for support and troubleshooting assistance.