When we last surveyed the Hermes Agent ecosystem on May 25, the numbers were already staggering: 165,000 GitHub stars, 27,300 forks, and 310+ community skills. Two weeks later, those numbers look quaint.
As of June 10, 2026, the NousResearch/hermes-agent repository has crossed 188,781 GitHub stars — a gain of over 23,000 stars in roughly two weeks. The Skills Hub now lists 90,881 skills across 12 registries, up from a few hundred. And Dealroom has officially designated Hermes Agent as the fastest-growing open-source agent framework of 2026.
"The open-source self-improving AI agent from Nous Research has now accumulated over 180,000 GitHub stars in under four months since its February 25 launch." — Ewan Mak, Medium
The surface-level story is growth. The deeper story is structural — the ecosystem is reorganizing itself around discoverability, curation, and plugin infrastructure that didn't exist two weeks ago.
The Skills Hub: From Hundreds to 90,881
The single most dramatic change is the Skills Hub. In late May, the community was sharing skills ad-hoc — GitHub gists, Reddit threads, Discord pins. The official Skills Hub changed that.
Now cataloguing 90,881 skills across 12 registries, the Hub auto-refreshes twice daily and includes contributions from NVIDIA/skills (added to trusted taps in v0.16.0), alongside community registries. A skill installed through the Hub isn't just a file download — it's a versioned, auditable artifact that Hermes's built-in Curator can grade, prune, and consolidate automatically.
The Skill ecosystem has also spawned its own meta-layer: lobehub.com now features a trending skills marketplace for Hermes Agent, updated daily, complete with trigger phrases and usage examples for each skill.
The Curation Layer: Awesome Lists and Atlas
Two weeks ago, finding community tools meant scrolling through Reddit or searching GitHub. Now there are three major discovery surfaces:
| Resource | Stars | Description |
|---|---|---|
| SamurAIGPT/awesome-hermes-agent | New | Comprehensive curated list — skills, plugins, tools, integrations, deployment templates. Updated 2 days ago. |
| 0xNyk/awesome-hermes-agent | Growing | Curated ecosystem guide with safety flows and evolution loops. Updated 1 week ago. |
| Hermes Atlas (ksimback/hermes-ecosystem) | Community map | 100+ tools mapped with live GitHub data and AI-powered search. Published "The State of Hermes — May 2026" report. |
This is the classic sign of a maturing open-source ecosystem: the project no longer needs to tell you what's available — the community builds the map.
New Plugin Infrastructure: Hermes LCM
The v0.16.0 Surface Release formally introduced the context engine plugin slot, and the community responded immediately. stephenschoettler/hermes-lcm — published just two days ago — is a Lossless Context Management plugin that replaces the default lossy summarization with a DAG-based context engine.
Instead of the agent periodically summarizing and discarding old messages (the standard approach that inevitably loses nuance), LCM builds a directed acyclic graph of conversation context. Every message, tool call, and result is preserved — the engine just prunes the graph traversal, not the data. For long-running agents that operate across days or weeks (cron jobs, infrastructure operators, research assistants), this is transformative.
The plugin architecture itself is significant: it means the community can now extend Hermes's internals at well-defined extension points rather than forking the core.
Mainstream Signals: "STARTUP EDITION" and Media Attention
The ecosystem growth is being noticed outside developer circles. A June 2026 "STARTUP EDITION" roundup on blog.mean.ceo frames Hermes Agent explicitly for founders and investors, highlighting the self-hosted, always-on architecture as a differentiator from SaaS agents. Multiple Medium publications have published deep-dive guides (Ewan Mak, ThePlanetTools), and a Substack guide now walks product designers, PMs, and CMOs through real-world use cases.
The Hermes Agent Challenge on DEV — covered in our last report — produced 14 community projects, but the ripple effects continue: several challenge participants have since published their projects as standalone open-source repos, feeding back into the ecosystem.
What's Powering This Growth?
Three structural factors are driving the acceleration:
The Surface Release (v0.16.0) landed June 5 with 874 commits, 542 merged PRs, and Hermes Desktop — a native app for macOS, Windows, and Linux. Desktop access lowered the barrier from "configure a CLI tool" to "download and start chatting."
The self-evolution loop — hermes-agent-self-evolution with DSPy + GEPA, covered June 8 — means Hermes can now improve its own skills, creating a feedback loop where more usage produces better skills, which attracts more users.
Plugin architecture maturity — the context engine plugin slot, MCP integration, and profile system mean the community can build at the edges without waiting for the core team. The hermes-lcm plugin went from concept to published repo in under a week.
What to Watch
The 188K star milestone will likely be 200K+ within weeks. But the more interesting trajectory is the Skills Hub: if it maintains even a fraction of its current growth rate, Hermes Agent will have the largest skill library of any open-source agent framework — by a wide margin.
The plugin ecosystem is still in its infancy. The context engine slot is the first of what will likely become a broader plugin API. Community projects like SkillClaw (a post-task evolution loop on top of Hermes's built-in skill creation) and the emerging fleet management tools suggest the next wave will be about coordinating multiple Hermes instances — not just running one.
For now, the numbers tell a clear story: Hermes Agent isn't just growing. It's developing the institutional infrastructure — skill registries, curated lists, plugin APIs, community maps — that transforms a popular project into a platform.
Cet article a été initialement publié sur The Agent Report.
