Your Claude subscription in Hermes, Pi or OpenCode
Hermes needs Max plus extra usage, Pi's sign-in bills as extra usage unless Claude Code carries the requests, OpenCode takes an API key. What Anthropic's terms say.
It depends on how the agent reaches Claude, and the three agents do it three different ways. Hermes signs in to Anthropic itself, works only on a Max plan with extra usage bought on top, and bills every request as extra usage, not against the plan’s included limits; Pro doesn’t work at all. Pi’s own Anthropic sign-in also bills as extra usage, per token; its makers publish an extension that sends Pi’s model requests through your installed Claude Code instead, and its README says those count against your subscription’s limits, while warning that the extension’s future is uncertain. OpenCode has no Claude sign-in: Claude models go in with an API key. As of October 23, 2026, Anthropic’s terms let you sign in to the unmodified Claude Code with your own plan, bar other developers from offering Claude login in their apps or storing your Claude credentials, and reserve enforcement “without prior notice.” Whether a given route is allowed is Anthropic’s call.
Three agents, three routes
| Agent and route | Plans | What the usage draws on |
|---|---|---|
Claude Code, signed in with your plan | Pro, Max, Team | Your plan’s limits; the use the plans are built for |
Hermes, Anthropic sign-in ( hermes model) | Max with extra usage bought | Extra usage only; the included Max allowance is untouched |
Pi, /login with Anthropic | A Claude plan with extra usage | Extra usage, billed per token, by Pi’s own warning |
Pi with the pi-claude-code extension | Pro or Max, through Claude Code | Your subscription’s quota and rate limits, by its README; Anthropic’s call, below |
OpenCode 2 | None | An Anthropic API key, billed per token |
Any of them with an Anthropic API key | None needed | The key’s organization, per token; on Max or Team, that organization’s monthly API credits |
The split that decides the billing is whether the agent talks to Anthropic itself with your sign-in, or hands the model request to Claude Code, Anthropic’s own program, running under your own login. OpenClaw offers both shapes, and the OpenClaw and Claude subscription guide sets its two routes beside the same terms.
Hermes: Max plus extra usage, or an API key
Hermes’s provider docs are plain about the limits of its Anthropic sign-in. When you authenticate through hermes model, then Anthropic OAuth, “Hermes routes as Claude Code against your Anthropic account,” and “It only works if you’re on a Claude Max plan and have purchased extra usage credits.” The base Max allowance is not consumed; only the extra usage you have added is. The docs’ own line on the surprise: “All Hermes usage bills as ‘extra usage’ even while your included Max allowance sits untouched.” And on Pro: “Pro looks like it should work; it doesn’t.”
What running out looks like is an error from Anthropic, not from Hermes, reported against Hermes v0.10.0 in a GitHub issue by a Max subscriber whose plan still had room:
HTTP 400: You're out of extra usage. Add more at claude.ai/settings/usage and keep going.The fix is more extra usage, or a different credential. Hermes’s docs send anyone without Max and extra usage to an ANTHROPIC_API_KEY, billed per token to that key’s organization. On the sign-in itself, Hermes prefers Claude Code’s own credential store over copying the token into ~/.hermes/.env, keeps sign-ins of its own in a credential pool that take precedence, and reads Claude Code’s login only as a fallback, which auth.adopt_external_logins: false turns off. Installing Hermes and choosing a provider are in the Hermes install guide.
Pi: two routes with opposite billing
Pi’s /login offers Anthropic, and once you use it, Pi shows: “Anthropic subscription auth is active. Third-party harness usage draws from extra usage and is billed per token, not your Claude plan limits.” That is the text of Pi 1.2.0; nothing in Pi’s changelog through 1.4.0, released October 21, 2026, changes the sign-in.
The other route is an extension from Pi’s makers, pi-claude-code, which “lets you use your Claude Pro or Max subscription in Pi by routing all Claude model requests through the official Claude Code CLI,” in its non-interactive mode, claude -p. Pi keeps the history, approvals, compaction and tools; Claude Code only talks to the model and never runs tools itself. It needs Claude Code installed and signed in, uses whatever login Claude Code has, and doesn’t read, copy or store those credentials. Its README installs it with:
npm install -g @anthropic-ai/claude-code
pi install https://github.com/earendil-works/pi-claude-code“Requests made through this extension count against your Claude subscription’s usage quota and rate limits.” The same README opens its caveats with “The future of this extension is uncertain”: Anthropic announced in May 2026 that claude -p and Agent SDK usage would move to a separate credit billed at API rates, the change is on hold, and if it lands the extension “may cost more, stop drawing from your subscription, or stop working altogether.” Everything else Pi adds through packages is in the Pi extensions guide.
OpenCode: an API key only
OpenCode used to ship a plugin that signed in with a Claude account. Version 1.3.0 (March 22, 2026) removed it, and OpenCode’s providers page now says: “There are plugins that allow you to use your Claude Pro/Max models with OpenCode. Anthropic explicitly prohibits this.” OpenCode 2 offers Anthropic by API key or ANTHROPIC_API_KEY and nothing else. OpenCode’s home page still says a Claude plan works; the OpenCode pricing guide sets the pages that disagree side by side, with the subscriptions that do sign in, ChatGPT among them.
What Anthropic’s pages say
Three of Anthropic’s pages bear on this, read on October 23, 2026:
- The terms for Claude Code. Anthropic’s legal and compliance page says subscription sign-in is “designed to support ordinary use of Claude Code and other native Anthropic applications,” that “Anthropic does not permit third-party developers to offer Claude.ai login into their own applications, or to route requests through Free, Pro, or Max plan credentials on behalf of their users,” and that developers “may not collect, store, or intermediate Claude.ai credentials or session tokens.” It does not prevent “an end user from signing in to the unmodified Claude Code binary with their own Claude subscription.” Anthropic “may do so without prior notice” when it enforces these restrictions.
- The help article on other tools. Anthropic’s article on logging in calls an API key the preferred way for third-party software, open-source projects included. It says Anthropic “may at its discretion allow paid subscribers who have enabled usage credits to use certain third-party tools,” drawing that use “from usage credits rather than subscription limits,” and it prohibits third-party tools that “misrepresent their identity to Anthropic’s servers” or “attempt to route third-party traffic against subscription limits.” It names no tool.
- The article on the Agent SDK and your plan. In its update of October 7, 2026: “You can still use the Claude Agent SDK, claude -p, and third-party apps with your subscription limits.” It doesn’t say which apps it means.
Set beside the routes above. Hermes’s and Pi’s own sign-ins are a Claude login offered inside another developer’s app, and each has that app store your subscription token, which Anthropic’s terms bar third-party developers from doing; Hermes’s docs also say it “routes as Claude Code.” Both bill as extra usage, the drawing on usage credits that the help article says Anthropic may allow at its discretion. The pi-claude-code route leaves the login inside your own Claude Code running claude -p, which the Agent SDK article lists, while sending another harness’s requests against your plan’s limits, which the help article names among what it prohibits. Which side of these lines a given route falls on is Anthropic’s call, and the plans’ advertised limits “assume ordinary, individual usage of Claude Code and the Agent SDK.” What the same terms mean for Claude Code on a server of your own is in the guide to your plan on a server.
On Max, an API key is partly prepaid
Since October 7, 2026, Max and Team plans include monthly API credits: $100 on Max 5x, $200 on Max 20x, and on Team $20 a Standard seat and $100 a Premium seat, pooled and capped at $500. You claim them by linking one Claude Console organization to the plan, and “Everyone with an API key in the linked organization draws from the same balance.” They cover the Claude API and the Agent SDK, not interactive Claude Code, and unused credits expire at the end of each billing cycle. So on Max, an Anthropic API key from that organization, put into Hermes, Pi or OpenCode, draws on the month’s credits, by the route Anthropic’s help article calls the preferred one for other tools. What each plan includes for Claude Code itself is in the Claude plans guide.