AlexAI-MCP/hermes-CCC
Hermes Agent ported to Claude Code Channel — 46 native skills, no OAuth, no external process
hermes-CCC is a port of the Hermes Agent framework specifically designed for the Claude Code native skill system. It integrates the Hermes brain architecture directly into the Claude Code execution context, eliminating the need for external processes or OAuth authentication. The project implements 46 core skills across categories such as MLOps, software development, and research, including specialized tools for RL training and vector database management. This allows developers to utilize Hermes-style memory, personas, and task routing while remaining within a unified terminal environment.
- Port of Hermes Agent brain and 46 core skills to CCC
- Runs natively in Claude Code without separate processes or OAuth
- Includes MLOps tools for vLLM, GGUF inference, and RL training
full readme from github
hermes-CCC
Hermes Agent, faithfully ported to Claude Code Channel (CCC).
hermes-CCC brings the power of Hermes Agent by NousResearch into Claude Code's native skill system — no separate process, no OAuth, no lock-in. Everything runs inside Claude Code.
Built by AlexAI-MCP.
What's Ported
| Hermes Component | hermes-CCC Equivalent |
|---|---|
agent/ brain |
/hermes-route, /hermes-memory, /hermes-skill, /hermes-traj, /hermes-persona, /hermes-compress, /hermes-search, /hermes-insights |
| Honcho user modeling | /honcho |
skills/ (402) |
46 core skills across 10 categories |
tools/ (68) |
Claude Code native tools + MCP servers |
gateway/ platforms |
Discord/Telegram via cc-channel plugin |
cron/ |
CronCreate tool |
environments/ |
Claude Code execution context |
plugins/memory/ |
cc-channel-mem + auto-memory |
Install
git clone https://github.com/AlexAI-MCP/hermes-CCC
cd hermes-CCC
chmod +x install.sh && ./install.sh
Windows (PowerShell):
git clone https://github.com/AlexAI-MCP/hermes-CCC
cd hermes-CCC
.\install.ps1
One-liner (Unix):
curl -fsSL https://raw.githubusercontent.com/AlexAI-MCP/hermes-CCC/main/install.sh | bash
Restart Claude Code after install to activate all skills.
Skill Catalog (46 skills)
Core Brain (Hermes Identity)
| Skill | Invoke | Description |
|---|---|---|
| hermes-route | /hermes-route |
Analyze task complexity → recommend opus/sonnet/haiku |
| hermes-memory | /hermes-memory |
Memory prefetch/sync/nudge (memory_manager port) |
| hermes-skill | /hermes-skill |
Skill lifecycle: create, improve, audit |
| hermes-traj | /hermes-traj |
Log interaction trajectories for analysis/RL |
| hermes-persona | /hermes-persona |
Switch persona: researcher/coder/analyst/creative/advisor |
| hermes-compress | /hermes-compress |
Compress context, extract session state |
| hermes-search | /hermes-search |
Search past sessions, memory files, project history |
| hermes-insights | /hermes-insights |
Usage analytics and productivity insights |
| honcho | /honcho |
Cross-session user modeling (dialectic user profile) |
Software Development
| Skill | Invoke | Description |
|---|---|---|
| systematic-debugging | /systematic-debugging |
4-phase root cause investigation |
| test-driven-development | /test-driven-development |
Red-Green-Refactor TDD cycle |
| subagent-driven-development | /subagent-driven-development |
Parallel subagent workstreams |
| github-code-review | /github-code-review |
Full PR code review workflow |
| github-pr-workflow | /github-pr-workflow |
PR lifecycle automation |
| github-issues | /github-issues |
Issue management and triage |
MLOps / AI
| Skill | Invoke | Description |
|---|---|---|
| vllm | /vllm |
OpenAI-compatible LLM inference server |
| llama-cpp | /llama-cpp |
Local GGUF inference (CPU+GPU) |
| whisper | /whisper |
Speech recognition and transcription |
| stable-diffusion | /stable-diffusion |
Text-to-image with diffusers |
| grpo-rl-training | /grpo-rl-training |
GRPO/RL fine-tuning with TRL |
| huggingface-hub | /huggingface-hub |
Model/dataset download, upload, search |
| flash-attention | /flash-attention |
2-4x attention speedup, 10-20x memory savings |
| instructor | /instructor |
Structured Pydantic outputs from any LLM |
| chroma | /chroma |
Open-source vector DB for local RAG |
| qdrant | /qdrant |
Production vector search engine |
| faiss | /faiss |
Ultra-fast in-memory vector similarity search |
| pinecone | /pinecone |
Managed cloud vector database |
Research
| Skill | Invoke | Description |
|---|---|---|
| arxiv | /arxiv |
arXiv paper search and download |
| research-paper-writing | /research-paper-writing |
Academic paper workflow with LaTeX |
| blogwatcher | /blogwatcher |
Monitor RSS feeds and blogs |
| duckduckgo-search | /duckduckgo-search |
Free web search, no API key needed |
| polymarket | /polymarket |
Prediction market probability data |
Productivity
| Skill | Invoke | Description |
|---|---|---|
| google-workspace | /google-workspace |
Gmail, Drive, Sheets, Calendar automation |
| linear | /linear |
Linear issue and project management |
| jupyter-live-kernel | /jupyter-live-kernel |
Notebook execution and kernel management |
Creative
| Skill | Invoke | Description |
|---|---|---|
| excalidraw | /excalidraw |
Diagrams and whiteboard sketches via MCP |
| manim-video | /manim-video |
Mathematical animations with Manim |
Infrastructure
| Skill | Invoke | Description |
|---|---|---|
| docker-management | /docker-management |
Container, image, Compose management |
| native-mcp | /native-mcp |
MCP server integration in Claude Code |
| mcporter | /mcporter |
Convert any CLI tool into an MCP server |
Security / OSINT
| Skill | Invoke | Description |
|---|---|---|
| sherlock | /sherlock |
Username search across 400+ social networks |
| oss-forensics | /oss-forensics |
Supply chain risk and dependency analysis |
| one-password | /one-password |
1Password CLI secret management |
Blockchain
| Skill | Invoke | Description |
|---|---|---|
| base-blockchain | /base-blockchain |
Base (Ethereum L2) on-chain data queries |
| solana | /solana |
Solana blockchain data via JSON RPC |
Media
| Skill | Invoke | Description |
|---|---|---|
| youtube-content | /youtube-content |
Download, transcript, and research YouTube |
Relationship to Hermes Agent
This is a Claude Code-native reinterpretation of Hermes operational patterns, not a byte-for-byte clone. The key differences:
- No separate process — everything runs inside Claude Code
- No OAuth — uses Claude Code's native auth
- No gateway — uses existing Discord/Telegram MCP plugins
- No model dependency — works with any Claude model
See docs/migration-guide.md for migration guidance from Hermes Agent.
Package Layout
hermes-CCC/
├── README.md
├── install.sh ← Unix/WSL installer
├── install.ps1 ← Windows PowerShell installer
├── CLAUDE.md ← Maintainer notes
├── LICENSE ← MIT
├── docs/
│ ├── migration-guide.md
│ └── tool-mapping.md
├── .github/
│ └── workflows/
│ └── validate.yml
└── skills/ ← 46 skill directories
├── hermes-route/
├── hermes-memory/
└── ...
Contributors
| Contributor | Role |
|---|---|
| AlexAI-MCP | Creator & Maintainer |
| Claude (Anthropic) | Skill design, porting, and architecture |
| Codex (OpenAI) | Parallel skill generation and build automation |
Acknowledgements
hermes-CCC is built on top of the excellent work done by the NousResearch team on Hermes Agent. The original framework's design — its brain architecture, skill system, memory management, and platform integrations — served as the direct blueprint for this port.
A sincere thank you to the Hermes Agent community and all contributors who made the original project what it is. This package exists because of your work.
hermes-CCC is an independent port. Not affiliated with NousResearch.
License
MIT © 2026 AlexAI-MCP