Hermes proposes a sign-up-free entry tier, but inference is not live

Hermes opened a September 7 proposal to make a fresh installation useful before its owner finds an API key. The proposed Nous free tier combines one model, nous/welcome, with managed connectors and an eventual account-upgrade flow. This is an onboarding direction, not an available free inference service: the inspected pull request remains open and explicitly says every chat scenario ended in HTTP 401 because welcome mode was disabled on the server.

The facts

- The proposal gives an installation both managed connectors and a single welcome model without an initial sign-up step. - An existing provider key still takes precedence for inference, while the proposed Nous identity supplies access to connectors. - The proposed nous.guest setting defaults to true; setting it false prevents free-tier setup and use for both inference and connectors. - Sibling profiles reuse an identity from a shared store rather than intentionally creating a separate identity for each profile. - The author reports successful identity creation and token exchange against a local account service, but no successful model reply. - Desktop picker and status support, messaging-gateway upgrade flow and the upgrade command's final name remain follow-up work.

Why it matters

The adoption barrier is not just the model bill. New operators often have to connect inference, browser access and third-party tools before the first useful task. Combining that bootstrap could make Hermes easier to evaluate while keeping another vendor's model in place. It also moves part of first-run behavior from local configuration to a hosted identity service. That distinction matters to teams choosing self-hosting for predictable network behavior: an account-free interface is not the same thing as an account-free architecture, and a free model route is not a promise of arbitrary model access.

Current

Inspected on 2026-09-08. The Hermes Agent stable-release baseline is v2026.9.7 published 2026-09-07T22:17:01Z. The main source was open when captured. Release metadata is a version boundary, not evidence that an open proposal has shipped.

Evidence

The primary source is NousResearch/hermes-agent PR #105258 (https://github.com/NousResearch/hermes-agent/pull/105258). Supporting context comes from Hermes Agent — current project README (https://raw.githubusercontent.com/NousResearch/hermes-agent/main/README.md); Hermes — official documentation (https://hermes-agent.nousresearch.com/docs/). The linked records were inspected directly; related project records are not independent confirmations.

Operator take

Watch for a release and a genuine first-answer receipt before describing this as launched. A useful acceptance matrix would cover a clean install, an existing external provider, the explicit opt-out, and an account upgrade that retains already-connected tools. The proposed route-based model pinning deserves attention because a credential pool may contain paid and welcome credentials together; the endpoint selected for the request must determine which model can actually run. The source also distinguishes sequential profile reuse from concurrent identity creation. Its locking design is not a substitute for testing simultaneous starts. No installation or account changes are warranted merely because this proposal exists.

Caveat

This is open implementation work, not a pricing commitment or launch announcement. Test claims belong to the PR author; the welcome server, browser consent completion and concurrent live identity creation were not proven in the reported acceptance run.