Amp Opens Up Claude Pro and Max Subscriptions Through a Claude Code Mode

Amp, Sourcegraph's coding agent, now has a Claude Code mode. You can use it with a Claude Pro or Max subscription, which means spending an Anthropic plan you have already paid for. The feature is free and available to all users, and the announcement is dated October 10, 2026. To enable it, create a new thread and select Mode, then Claude Code. If your subscription is not yet linked, Amp asks you to link it.

What changed in the interface

Amp switches between working modes with a "dial" (the dial), and Claude Code is now one of its positions. In the screenshot from the announcement, this mode is selected, a Pro/Max plan is connected, the model is listed as Opus 5.5, and the effort level is set to High. In other words, reasoning settings are available here just as they are in Amp's other modes.

The main practical benefit is that you do not have to pay for model access twice. Anyone who already has a Claude subscription for terminal-based coding does not need a separate budget for Amp. The announcement does not say how plan limits are consumed. It is reasonable to expect that they are shared with the rest of your subscription usage, but it is best to check the documentation.

Under the hood: someone else's agent in Amp's shell

The new mode runs the Claude Agent SDK, not Amp's own agent. Anthropic describes this way of using a subscription in a separate help-center article. That detail matters: this is a sanctioned route, not a workaround for restrictions.

Sourcegraph emphasizes that it managed to preserve the familiar Amp environment around the SDK. The following all work:

• orbs and runners (Amp's own terms, not explained in the announcement);
• thread sharing;
• portals;
• multiplayer, meaning collaboration within a single thread;
• messaging between threads;
• orchestration of multiple threads.

This is not a trivial task. Much of it depends on how Amp's own agent is built, and here a third-party SDK drives the loop. So the claim that "everything works" is worth testing on real tasks, especially in setups involving several threads.

Key details

• Access: free, for all Amp users.
• Eligible plans: Claude Pro and Claude Max.
• How to enable: New Thread, Mode, Claude Code, then link your subscription.
• Engine: the Claude Agent SDK instead of Amp's own agent.
• Model in the screenshot: Opus 5.5, effort High.
• Preserved features: sharing, portals, multiplayer, thread orchestration.

How to try it and what to watch for

Start with a simple task in a new thread: select Claude Code, link your subscription, and compare the result with Amp's regular mode on the same repository. Agents behave differently: each has its own system prompts, toolset, and editing style. That is why a fair comparison is only possible on identical tasks.

Teams should check three things in advance. First, how the subscription is shared among people, since Pro and Max plans are personal. Second, how the mode behaves in orchestration, where one thread launches others. Third, whether the plan's limits hold up during long sessions at high effort. Details are in the Amp documentation section on using a Claude subscription (the-dial, Use your Claude subscription). For anyone already living in the Anthropic ecosystem, this is the shortest way to try Amp without new bills.