Hermes Agent v0.14.0 released. Subscription users can call major services without an API key

NousResearch released Hermes Agent v0.14.0 on May 16, codenamed “The Foundation Release”. The three core updates are the addition of hermes proxy instructions, which can convert web-based authorizations for Claude Pro, ChatGPT Pro, and other services into standard OpenAI-format local API endpoint calls.

hermes proxy: confirmed feature specifications

According to NousResearch’s official announcement confirming：

Function definition: Convert web-based OAuth authentication for Claude Pro, ChatGPT Pro, and SuperGrok into standard OpenAI-format local API endpoint calls

Supported developer tools: Codex, Aider, Cline, Continue, etc. can call directly

Usage conditions: Hold existing web subscription authorizations; no additional API Key application is required

Billing method: Calling models via this feature does not require paying API usage fees; it is calculated only based on existing subscription fees

xAI/Grok native upgrade: confirmed technical specifications

According to NousResearch’s official announcement confirming：

grok-4.3 context window: expanded to 1 million tokens

SuperGrok users: no need to apply for an API Key; you can call the model directly

x_search tool: changed from an external plugin to a native built-in tool, connecting the X (formerly Twitter) OAuth authorization flow, enabling native integration between Hermes Agent and the X platform

LSP semantic diagnostics: confirmation differences from v0.13.0

According to NousResearch’s official announcement confirming：

Trigger conditions: Each time the AI calls write_file or patch, the system automatically runs a round of LSP semantic analysis

Detectable types: type errors, undefined symbols, missing imports

Difference from v0.13.0: v0.13.0 provides basic Python/JSON/YAML syntax-format checking (linting); v0.14.0’s LSP analysis includes a semantic layer, allowing detection of program errors that syntax-level checks cannot capture

Other confirmed features

According to NousResearch v0.14.0 release notes confirming：

handoff instruction: supports seamless model switching with context, without needing to resend the conversation history

Microsoft Teams integration: completes end-to-end integration, covering Graph authorization, webhook listening, pipeline execution, and message sending

Shorter cold start time: heavy dependencies such as audio, video, and interactions on some platforms are changed to lazy loading; the official confirmed that cold start time is reduced by 19 seconds

FAQ

What is Hermes Agent?

An open-source, multi-platform AI agent framework developed by NousResearch. It supports multiple large language model providers and can perform tasks such as program code editing, web search, and platform integration. v0.14.0 includes 808 commits, with 215 community contributors.

Which existing subscribed users can use hermes proxy directly?

According to NousResearch confirmation, hermes proxy supports existing subscribers of Claude Pro, ChatGPT Pro, and SuperGrok, converting their web-based OAuth authorizations into local OpenAI-format endpoints, allowing developer tools such as Codex and Aider to call them directly—without needing to apply for an API Key or pay API usage fees.

What are the practical differences between LSP semantic diagnostics and the older linting?

According to NousResearch’s feature description, linting only checks code formatting. LSP analysis includes a semantic layer, enabling immediate detection—after the AI writes files—of issues such as type errors, undefined symbols, and missing imports, which are typically not detectable at the formatting level.