Hermes Agent is an open-source, self-improving AI agent built by Nous Research. Unlike static AI assistants, Hermes has a closed learning loop: it creates skills from complex tasks, improves them during use, and builds a deepening model of who you are across sessions.
Key Features:
- Self-Improving Learning Loop: Autonomously creates and refines skills from experience. Searches past conversations via FTS5 with LLM summarization for cross-session recall.
- Model-Agnostic: Works with OpenRouter (200+ models), Nous Portal, NVIDIA NIM, OpenAI, Hugging Face, and any custom endpoint. Switch providers with
hermes model— no code changes. - Multi-Platform Presence: Runs in a full TUI terminal, and via a gateway on Telegram, Discord, Slack, WhatsApp, Signal, and Email — all from one process.
- Runs Anywhere: Six terminal backends — local, Docker, SSH, Daytona, Singularity, and Modal. Serverless persistence via Daytona/Modal means near-zero cost when idle.
- Scheduled Automations: Built-in cron scheduler for daily reports, nightly backups, and weekly audits in natural language.
- Parallel Subagents: Spawns isolated subagents for parallel workstreams. Python scripts can call tools via RPC.
- Research-Ready: Batch trajectory generation, Atropos RL environments, and trajectory compression for training tool-calling models.
Installation:
curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bashWorks on Linux, macOS, WSL2, and Android via Termux. Compatible with the agentskills.io open standard.
