Hermes Agent launches official marketplace: MCP, skills, and plugins in one place
Nous Research, the team behind the Hermes AI agent, has launched an official extension marketplace. Instead of searching GitHub and chats for MCP servers and skills, users can now find everything in a single catalog organized by category. For people building agents for their own tasks, this saves time on discovery and initial vetting. The available plugins include one that connects the agent through a Claude subscription rather than the API.
AI-processed from TG: @neuraldvig (Нейродвиж); edited by Hamidun News
Anyone who has built a useful AI agent knows the ritual. You need an email connector, a skill for working with spreadsheets, and something for the browser. Then comes the search through repositories, GitHub collections, and chats where someone recommends “this thing; it seems to work.” The team behind Hermes, the agent from Nous Research, decided to close this gap by launching an official extension marketplace.
What is inside
The catalog contains MCP connectors, skills, and other plugins organized by category so users can quickly find what they need. If these terms are unfamiliar, the distinction is straightforward. MCP is a protocol through which an agent accesses external services such as a calendar, database, or task tracker. A skill is a ready-made set of instructions and workflows for a common task, such as “prepare a weekly report.” A plugin extends the capabilities of the agent itself.
Consider a typical task: an agent needs to collect requests from Notion and create corresponding tasks in a tracker. Previously, the process looked like this: find an MCP server, determine whether the project is still maintained, work through the setup, and hope it does not break after an update. The new flow is different: open the relevant category, find a connector, and connect it. It is not magic, but in everyday work the difference between “half an hour of searching” and “five minutes to install” is substantial.
The original post does not provide a complete list of available items, so it is too early to judge the quality or quantity of the extensions. Users should inspect the catalog themselves.
A plugin for Claude subscriptions
The most discussed item is a plugin that makes it possible to use the agent through a Claude subscription rather than paying for API usage. The appeal is clear. Agents consume many tokens: they operate in loops, reread context, and call tools. With token-based billing, costs can rise unexpectedly, while a subscription has a fixed price. For an enthusiast running an agent on personal tasks, this looks attractive.
There are several caveats, however. First, a provider's terms for subscription and API use may differ, and methods that bypass the API may violate the rules or stop working after an update. Second, subscriptions have limits, and a demanding agent workflow will reach them faster than an ordinary chat session. Third, teams and production systems need predictability: an SLA, clearly defined limits, and support. The API provides these; a consumer subscription does not. The option is therefore interesting for personal experiments, but it should not underpin a business process.
Limitations and practical takeaways
The main point to remember is that a marketplace distributes third-party code. An MCP connector gains access to your data and may perform actions on your behalf. Add prompt injection, in which malicious instructions are hidden in text processed by the agent, and excessive permissions granted “just in case.” The catalog's official status does not itself guarantee that every plugin has undergone a thorough review. Follow a few simple rules:
- read the source code before installation, especially for connectors that can write data;
- grant minimum permissions and use separate tokens rather than all-purpose keys;
- test first in a sandbox or with test accounts;
- pin versions so an update cannot unexpectedly change the agent's behavior.
What does this mean for products and founders? Agent ecosystems are moving toward the extension-store model already seen with browsers and IDEs. Whoever secures an early position in the catalog of a popular agent gains a low-cost distribution channel. If your service has an API, an MCP connector or skill for it can put the product in front of users who already work with the agent. For developers, the takeaway is simpler: before writing custom integration code, check the catalog because what you need may already exist. Treat extension installation as seriously as adding a dependency to a project.
- Nov 8, 2026US Elections Wipe $3.2 Trillion Off Tech Market and Raise Pressure on Data Centers
- Nov 8, 2026Claude Detects Implanted Thoughts 20% of the Time but Cannot Name Them
- Nov 8, 2026Investors Push Anthropic to Cut Its $1.5 Trillion Valuation Ahead of IPO
- Nov 8, 2026Anthropic investors rethink IPO valuation: it could fall to $1.2 trillion
Key Questions Answered
What is inside?
The catalog contains MCP connectors, skills, and other plugins organized by category so users can quickly find what they need. If these terms are unfamiliar, the distinction is straightforward. MCP is a protocol through which an agent accesses external services such as a calendar, database, or task tracker. A skill is a ready-made set of instructions and workflows for a common task, such as “prepare a weekly report.” A plugin extends the capabilities of the agent itself.
What are the key details on A plugin for Claude subscriptions?
The most discussed item is a plugin that makes it possible to use the agent through a Claude subscription rather than paying for API usage. The appeal is clear. Agents consume many tokens: they operate in loops, reread context, and call tools. With token-based billing, costs can rise unexpectedly, while a subscription has a fixed price. For an enthusiast running an agent on personal tasks, this looks attractive. There are several caveats, however.
What are the key details on Limitations and practical takeaways?
The main point to remember is that a marketplace distributes third-party code. An MCP connector gains access to your data and may perform actions on your behalf. Add prompt injection, in which malicious instructions are hidden in text processed by the agent, and excessive permissions granted “just in case.” The catalog's official status does not itself guarantee that every plugin has undergone a thorough review.
Build & Ship Production AI Agents Without Boilerplate Code
Start your free 72-hour trial at Hamidun Academy (academy.hamidun.com) — interactive lessons on Claude Code & Cursor, live Tech Radar, and production agent templates.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.