๐ UPDATE โ May 16, 2026
xAI has expanded Hermes Agent with two significant additions: X Premium subscriptions are now supported directly within the agent, and Hermes Agent has gained the ability to search X posts in real time. This deepens the integration between Grok's agent layer and the X platform, giving users access to live social data alongside existing AI capabilities. The update was announced officially by @xai on May 16, 2026.
![]()
โ @xai on X ยท May 16, 2026
xAI just flipped a switch that lets Grok subscribers use their existing subscription inside NousResearch's Hermes Agent โ no new API key required. The integration went live on May 15, 2026, and brings Grok's full capability stack directly into an open-source agent designed to run persistently and build memory across sessions.

What exactly is the NousResearch Hermes Agent?
Hermes Agent is an open-source, self-improving AI agent built by NousResearch and launched in February 2026. Unlike a standard chatbot, it runs persistently โ meaning it retains long-term memory across sessions rather than starting fresh each conversation. It also connects to messaging platforms including WhatsApp, Discord, Telegram, and Signal. The version current at the time of this integration is 0.13.0, dubbed "The Tenacity Release," which shipped on May 7, 2026.
What Grok capabilities are available inside Hermes?
The integration isn't limited to basic text chat. According to xAI, Grok subscribers get access to the following inside Hermes Agent:
- Grok 4.3 โ text conversations and advanced reasoning
- Grok Text-to-Speech โ spoken responses generated from text
- Grok Imagine โ image and video creation
A single OAuth bearer token covers all of these services within Hermes, so there's no need to authenticate separately for each capability.
Who can use this โ do you need a specific Grok tier?
All Grok subscription tiers are supported. xAI hasn't restricted this integration to premium plans, which means any paying Grok subscriber can connect their account to Hermes Agent from day one.
How do you actually connect Grok to Hermes Agent?
Setup is handled through a standard OAuth flow โ no manual API key configuration needed. When signing into Hermes Agent, users get a browser-based OAuth 2.0 PKCE login option via accounts.x.ai. That's it. The integration reuses Hermes's existing codex_responses adapter under the hood, so it supports reasoning, tool-calling, streaming, and prompt caching without any additional configuration on the user's end.
What is prompt caching and why does it matter here?
Prompt caching reduces both latency and cost in multi-turn conversations by reusing previously processed context instead of recomputing it from scratch each time. Hermes Agent automatically activates Grok's prompt caching by sending the x-grok-conv-id header with API requests whenever it detects an xAI endpoint. For an agent designed to maintain long-running, memory-rich sessions, this is a meaningful efficiency gain โ conversations that build on prior context won't be as computationally expensive to sustain.
Why does this matter beyond the technical details?
The pairing of Grok with a persistent, self-improving agent is a meaningful step beyond what a standalone AI subscription typically offers. Most AI tools reset context at the end of a session. Hermes is built to accumulate knowledge over time, and plugging Grok's reasoning and multimodal capabilities into that architecture gives subscribers a more capable long-running assistant โ one that can generate images, speak responses, and reason through complex tasks without losing the thread of prior interactions.
For Grok subscribers who have been looking for more ways to put their subscription to work beyond the standard X interface, Hermes Agent now represents a genuinely different use case โ persistent, cross-platform, and multimodal from the start.

Sarah focuses on Tesla Energy, SpaceX missions, and the broader Musk AI portfolio. Former data analyst in clean energy. Based in San Francisco.
Sources verified at publish time. Spotted an inaccuracy? Email editorial@basenor.com.







