Nous Research's Hermes Agent Ships a Built-in Tool Store Without Writing YAML

Hermes Agent now ships a curated, one-click MCP catalog — letting you install reviewed integrations like GitHub, Linear, and n8n without touching a config file.

·
·
·
Nous Research's Hermes Agent Ships a Built-in Tool Store Without Writing YAML
Read6 min
TypeNews
  • Hermes Agent now ships a built-in MCP Catalog with an interactive terminal picker for one-click tool server installs.
  • Catalog entries (GitHub, Linear, n8n, and more) are Nous-reviewed via PR merge — no unvetted community submissions.
  • Install flow handles credentials automatically: API keys go to ~/.hermes/.env, OAuth servers open a browser flow with full PKCE support.
  • Per-server tool filtering lets you whitelist or blacklist individual tools, cutting token overhead and reducing attack surface.
  • Hermes can also run as an MCP server itself, exposing its messaging layer to Claude Code, Cursor, or any MCP client.
  • The update shipped alongside a native desktop app and a full browser-based admin panel in the v0.15 release.

Hermes Agent, the open-source self-improving agent from Nous Research, just shipped a built-in MCP Catalog , a curated, interactive store of pre-reviewed tool integrations you can install with a single keypress. If you've ever wanted your agent to manage GitHub issues, trigger n8n workflows, or query Linear tickets without writing a line of YAML, this is the feature that makes it trivial.

The problem it's solving

MCP (Model Context Protocol) is the emerging open standard that lets AI agents connect to external tool servers , think GitHub, databases, file systems, Stripe, and internal APIs , without the agent developer having to build native integrations from scratch. The protocol has been gaining momentum fast, but the setup experience has been rough: find a server, read its docs, figure out the right command and args, handle OAuth, manage credentials. Every integration was a manual config surgery job.

Hermes now ships a curated catalog of MCP servers that Nous staff has reviewed and merged. They're disabled by default , install only what you actually want. The catalog is surfaced through an interactive terminal picker, and the whole flow from "I want this tool" to "it's wired up and working" is a few keypresses.

How it works

The catalog is accessed via three commands:

cmake
hermes mcp              # interactive picker (default)
hermes mcp catalog      # plain-text list, scriptable
hermes mcp install n8n  # install a catalog entry by name

Hitting Enter on a row walks you through installing (including any required credentials), and lets you enable, disable, or uninstall. The picker shows each entry with a live status label , available, enabled, or installed (disabled) , so you always know what's active.

After credentials are configured, Hermes does something particularly useful: it probes the MCP server to list every tool it exposes and presents a checklist, letting you select exactly which tools you want the agent to see. You can whitelist only create_issue and list_issues from GitHub, for example, and leave out anything destructive.

The trust model

This is where the catalog design gets interesting. Catalog entries are stored under optional-mcps/ in the hermes-agent repo , presence in that directory means Nous approval. There is no community submission tier; entries are added by merging a PR. That's a deliberate choice: it keeps the catalog small and audited rather than turning into an unreviewed marketplace.

Manifests are gated by PR review into the hermes-agent repo, so Nous has reviewed each entry before it shipped , but you should still read the manifest before installing, especially the source: field's repository, the install.bootstrap: commands, and any transport.command: invocation. The picker prints the upstream repo URL at install time so you can verify it before anything runs.

What's in the catalog

The catalog currently includes integrations across several categories. Examples visible in the docs include:

  • n8n , manage and inspect n8n automation workflows from Hermes
  • Linear , issue and project management via remote OAuth
  • GitHub , repo and PR tools via a local stdio server

Most hosted MCP servers , Linear, Sentry, Atlassian, Asana, Figma, Stripe , require OAuth 2.1 instead of a static bearer token. Setting auth: oauth has Hermes handle discovery, dynamic client registration, PKCE, token exchange, refresh, and step-up auth via the MCP Python SDK. For API-key-based servers, Hermes prompts at install time and writes the value to ~/.hermes/.env automatically.

Tool filtering and security controls

One of the more practical additions alongside the catalog is granular per-server tool filtering. You can whitelist or blacklist specific tools from any MCP server:

yaml
mcp_servers:
  github:
    command: "npx"
    args: ["-y", "@modelcontextprotocol/server-github"]
    env:
      GITHUB_PERSONAL_ACCESS_TOKEN: "***"
    tools:
      include: [list_issues, create_issue, update_issue]
      prompts: false
      resources: false

This matters for safety: if you're connecting Hermes to Stripe, you probably don't want the model to ever see delete_customer or refund_payment. The filtering is a first-class security control, not an afterthought. A Hermes deployment with five MCP servers and 34 tools shows average prompt sizes of 45,000 tokens per turn, with roughly 22,000 of those tokens , around 50% , being tool schema overhead alone. Filtering down to only the tools you need directly cuts that cost.

Hermes can also be an MCP server

The update isn't just about consuming MCP servers , Hermes can now also expose itself as one. This bidirectional MCP capability is architecturally interesting. A team running Claude as a primary reasoning layer can route specific tool calls or specialized tasks to a locally-running Hermes Agent instance via MCP, using Hermes as a specialized subagent rather than a standalone system.

Running hermes mcp serve starts a stdio MCP server that exposes 10 tools covering conversations, message history, event polling, and cross-platform message sending. This means Claude Code, Cursor, or any other MCP client can use Hermes as a bridge to Telegram, Discord, Slack, and the other messaging platforms it supports.

The bigger picture

Hermes is described as the self-improving AI agent built by Nous Research , the only agent with a built-in learning loop that creates skills from experience, improves them during use, nudges itself to persist knowledge, and builds a deepening model of who you are across sessions. The MCP catalog fits squarely into that philosophy: rather than Nous building every integration natively, the catalog lets the ecosystem's existing MCP servers plug in with a quality gate.

The v0.15 release that shipped the catalog was built across 100 PRs and 159 commits in a single week, and also introduced a native desktop app for macOS/Linux/Windows alongside a full browser-based administration panel that includes the MCP catalog, messaging channels, credentials, and webhooks. You no longer need to SSH into a server and edit YAML to wire up a new integration.

The repo currently sits at 183k stars and 29.8k forks on GitHub, and the project is MIT-licensed. To get started, install Hermes and run hermes mcp to open the catalog picker. To contribute a new MCP entry, open a PR against the optional-mcps/ directory in the repo , that's the only path to catalog inclusion.

Trending
  • No trending articles

Comments

avatar

Next Reads