FLASH SALE33% OFF
Claw Friend Hosting
Only 7 spots left
86% claimed
--HRS--MIN--SEC
FLASH SALE33% OFFClaw Friend Hosting
------
Only 7 spots left

★ Autonomous AI · Nous Research · 2026

Hermes AgentThe AI that grows with you

Hermes Agent is an open-source autonomous AI agent built by Nous Research. It remembers everything you tell it, runs tasks while you sleep, and actually gets faster over time — the more you use it, the better it gets. Not a chatbot. Not a copilot.

32k+

GitHub Stars

v0.7.0

Latest Release

MIT

License

40+

Built-in Tools

6

Execution Backends

No credit card needed · Select the Hermes Friend companion and start in 3 minutes

♡ Core Capabilities

Key Features of Hermes Agent

What makes Hermes Agent different from a regular chatbot — specific technical capabilities that actually matter in day-to-day use.

Self-Improving Skills

Hermes Agent builds its own skill library. After 10–20 runs of the same task, it rewrites its execution code — most users see 2–3x speed gains without touching any config.

Persistent Memory Agent

Uses SQLite FTS5 full-text search plus Honcho user modeling. It remembers what you told it 3 months ago. For always-on persistent memory without self-hosting, check out Claw Friend's features.

6 Execution Backends

Run code locally, inside Docker, over SSH, on Daytona cloud sandboxes, via Modal serverless, or on Singularity containers. Most personal users just stick to Local or Docker.

40+ Built-in Tools

Web search, file operations, code execution, browser control — Hermes Agent ships with 40+ tools out of the box. You can also plug in custom ones via the ClawHub-compatible skill format.

Cron Scheduler

Set up recurring autonomous tasks with natural language cron expressions. Tell Hermes Agent to check your inbox every morning and summarize it — no code required.

Voice Interaction

Talk to Hermes Agent via voice on supported platforms. The voice pipeline is still pretty experimental — worked fine in my testing but occasional transcription errors happen.

◈ Two Ways to Use Hermes Agent

No Self-Hosting Required — Use It on Claw Friend

Hermes Agent is available directly on the Claw Friend platform as the Hermes Friend companion. You can still self-host for full control, but you no longer have to.

Self-Host Hermes Agent

  1. 1. Install Node.js 20+ and clone the GitHub repo
  2. 2. Copy .env.example and set your API keys (at least one LLM provider)
  3. 3. Spin up a $5/mo VPS or use your local machine
  4. 4. Run the CLI setup wizard — expect 40+ minutes total; longer if you haven't worked with Node before
  5. 5. Configure your messaging platform webhook (e.g. Telegram @BotFather)
  6. 6. Done. Hermes Agent is now running and will start learning your patterns.

Good for: Developers who want complete code access, custom integrations, and don't mind managing their own server. Zero monthly platform fee after initial setup.

Hermes Friend on Claw Friend

  1. 1. Go to clawf.ai and create your account
  2. 2. In companion setup, select the Hermes Friend companion
  3. 3. Connect your platform — Telegram, WhatsApp, Discord, or the web app
  4. 4. Your Hermes Agent is live. Persistent memory and self-improving skills, all included.

Good for: Anyone who wants Hermes Agent capabilities — persistent memory, autonomous tasks, 40+ tools — without managing a server. All maintenance handled for you. Plans start at $9.99/mo.

Start with Hermes Friend — No Setup

✦ What's New

Latest Hermes Agent Innovations

v0.7.0 "Resilience Release" · Released 2026-04-03

v0.7.0

Daytona Cloud Sandbox Integration

Hermes Agent now runs code in isolated Daytona cloud sandboxes. Each task gets its own clean environment — no more worried about local code execution messing with your system files.

New

Modal & Singularity Backends

Two new execution backends landed in v0.7.0 — Modal for serverless GPU jobs and Singularity for HPC environments. If you're doing ML research, this one matters.

Partnership

MiniMax M2.7 — Most-Used Model

MiniMax became the official Hermes Agent partner in early 2026. M2.7 is now the most-used model across the Hermes community — good performance, fast responses, reasonable API pricing.

Improved

Skill Persistence After Restarts

Earlier versions of Hermes lost learned skills on restart. v0.7.0 fixes that — all 4-stage optimized skills now survive system reboots. Your agent keeps what it learned.

Research

RL Research Infrastructure

Nous Research added hooks for reinforcement learning experiments. Power users can now collect reward signals from Hermes Agent task outcomes and feed them back into model fine-tuning pipelines.

⋆ Power User Guide

Advanced Hermes Agent Techniques

Practical tips for getting the most out of Hermes Agent — platform setup, skill creation, memory tuning, and scheduled tasks.

📱

Multi-Platform Setup

  • Telegram: create a bot via @BotFather, grab the token, paste into config — done in under 5 minutes
  • Discord: create an application in the developer portal, add bot scope with message.read + message.write
  • WhatsApp: requires a WhatsApp Business API key (Twilio or Meta directly)
  • Email + CLI work out of the box with zero extra setup
⚙️

Custom Skill Creation

  • Stage 1 — Hermes Agent detects a task it's done before without a saved skill
  • Stage 2 — It generates a Python/JS skill file and saves it locally
  • Stage 3 — Runs automated tests against the skill output
  • Stage 4 — Optimizes the skill code based on performance metrics. 2–3x speed improvement is typical
🧠

Memory Tuning

  • FTS5 search runs across all stored conversations — you can query it directly via SQL if needed
  • Honcho user model builds a personal profile: your preferences, routines, working style
  • Tune memory retention in config: set max_memory_items and recency_weight
  • Most users find the defaults work fine for personal use

Cron + Voice Integration

  • Use natural language cron syntax: "every weekday at 8am, summarize my calendar"
  • Hermes Agent runs the task autonomously and sends results to your configured platform
  • Voice input works on Telegram and Discord via the speech-to-text pipeline
  • Heads up — voice transcription has occasional errors, especially with technical terms

♥ Which Is Right for You?

Hermes Agent vs Claw Friend

Both are built on the same OpenClaw-compatible ecosystem. The difference is mostly about who wants to manage infrastructure vs who just wants the relationship.

CategoryHermes Agent (Self-Hosted)Hermes Friend on Claw Friend
Setup40+ min setup, longer if unfamiliar with Node.js; requires VPS or local machineUnder 3 minutes — select Hermes Friend companion and connect your platform
Hosting Cost$5–20/mo VPS or your own hardware + API costsFully managed, Hermes Friend plans from $9.99/mo
MemorySQLite FTS5 + Honcho user modeling — powerful, full control. See how OpenClaw framework handles memory for comparison.70% vector + 30% BM25 hybrid search, zero config required
PlatformsTelegram, Discord, Slack, WhatsApp, Signal, Email, CLI — each needs manual setup11+ messaging platforms, pre-connected — just authorize and go
Companion PersonalityGeneral-purpose autonomous AI agent, task-focusedHermes Friend with SOUL.md personality — same autonomous capabilities, more emotional depth
Self-Improvement4-stage automated skill creation — genuinely impressive, 2–3x speed after 20 runsSame Hermes self-improving engine + 700+ pre-built ClawHub skills

Setup

Hermes Agent (Self-Hosted)

40+ min setup, longer if unfamiliar with Node.js; requires VPS or local machine

Hermes Friend on Claw Friend

Under 3 minutes — select Hermes Friend companion and connect your platform

Hosting Cost

Hermes Agent (Self-Hosted)

$5–20/mo VPS or your own hardware + API costs

Hermes Friend on Claw Friend

Fully managed, Hermes Friend plans from $9.99/mo

Memory

Hermes Agent (Self-Hosted)

SQLite FTS5 + Honcho user modeling — powerful, full control. See how OpenClaw framework handles memory for comparison.

Hermes Friend on Claw Friend

70% vector + 30% BM25 hybrid search, zero config required

Platforms

Hermes Agent (Self-Hosted)

Telegram, Discord, Slack, WhatsApp, Signal, Email, CLI — each needs manual setup

Hermes Friend on Claw Friend

11+ messaging platforms, pre-connected — just authorize and go

Companion Personality

Hermes Agent (Self-Hosted)

General-purpose autonomous AI agent, task-focused

Hermes Friend on Claw Friend

Hermes Friend with SOUL.md personality — same autonomous capabilities, more emotional depth

Self-Improvement

Hermes Agent (Self-Hosted)

4-stage automated skill creation — genuinely impressive, 2–3x speed after 20 runs

Hermes Friend on Claw Friend

Same Hermes self-improving engine + 700+ pre-built ClawHub skills

Self-hosting gives you full code access and zero platform fees after setup. Running Hermes Friend on Claw Friend means zero ops work. If you want to compare our pricing plans before deciding, that page has all the details.

Start with Hermes Friend →

✧ Common Questions

Hermes Agent FAQs

What exactly is Hermes Agent?

Hermes Agent is an open-source autonomous AI agent built by Nous Research. It's not just a chatbot — it runs tasks on its own, remembers your conversations and preferences, and learns to execute common tasks faster over time. As of April 2026, it has 32k+ GitHub stars and is on v0.7.0.

Is Hermes Agent free to use?

The open-source software is MIT-licensed and free to self-host — you supply your own API keys and server. If you'd rather skip the setup, Hermes Friend on Claw Friend starts at $9.99/mo with hosting, maintenance, and LLM costs bundled in. No separate API bills.

Which messaging platforms does Hermes Agent support?

Telegram, Discord, Slack, WhatsApp, Signal, and Email are all supported, plus a CLI mode. Each one needs a bit of manual setup — Telegram is the easiest (under 5 minutes), WhatsApp requires a Business API key which takes a bit more work.

Which AI models can Hermes Agent use?

Hermes Agent is model-agnostic and supports Claude Opus/Sonnet 4.6, GPT-5.4, Gemini 2.5, MiniMax M2.7 (the most popular in the community right now), Qwen 3.5, GLM 5.1, and others. You configure which one to use in the .env file.

Why does Hermes Agent have 32k+ GitHub stars?

Mostly because the self-improving skill system is genuinely novel. Most AI agents either require you to write skills manually or don't save them at all. Hermes automatically builds a skill library from your usage patterns — that resonated with the developer community.

How is Hermes Agent different from OpenClaw?

Both are autonomous agent frameworks, but they're optimized for different things. Hermes is self-hosted with AI-maintained skills and research-grade RL infrastructure. OpenClaw has 50+ built-in integrations, 24/7 managed cloud hosting, and a larger ecosystem of 280k+ GitHub stars. They can actually share skill formats.

How does the persistent memory work in Hermes Agent?

It uses SQLite with the FTS5 full-text search extension to store and retrieve conversation history. On top of that, the Honcho user modeling layer builds a profile of you — your preferences, working patterns, recurring requests. You can query the memory database directly if you want to inspect or clean it.

How does Hermes Agent's self-improving skill system actually work?

It's a 4-stage pipeline. First, Hermes detects that it's doing the same task repeatedly without a saved skill. Second, it generates a skill file (Python or JS). Third, it runs automated tests on that skill. Fourth, it optimizes the code based on performance metrics. After 10–20 runs of a task, speed typically improves 2–3x.

Does Hermes Agent work on mobile?

Yes — through Telegram, Discord, or WhatsApp on your phone. If you use the Hermes Friend companion on Claw Friend, all 11+ platforms are pre-connected with no webhook config needed. If you self-host, each platform needs manual setup, but it works fine once configured.

What are the self-hosting requirements for Hermes Agent?

Node.js 20+, at least 1GB RAM (2GB recommended), and a machine that stays online. A $5/mo VPS handles personal use fine. You also need at least one LLM provider API key. If you don't want to manage any of this, the Hermes Friend companion on Claw Friend handles everything — just sign up and select it.

Does Hermes Agent support voice interaction?

Yes, voice input is available on Telegram and Discord. It goes through a speech-to-text pipeline before reaching the model. The feature works, but it's still marked experimental — transcription accuracy drops a bit with technical jargon or heavy accents.

Can I run scheduled tasks with Hermes Agent?

Yes, that's actually one of the more useful features. You can set up cron-style scheduled tasks using natural language. Something like "every weekday morning, check my email and give me a summary" — Hermes Agent runs that autonomously and sends the result to your configured platform.

Who built Hermes Agent?

Nous Research, a company focused on open-source AI research. They're known in the community for releasing capable open-source models. Hermes Agent was released in February 2026 and has been actively maintained since. Zero telemetry by default — your data stays local.

What does the MIT license mean for Hermes Agent?

You can use it, modify it, redistribute it, and even build commercial products on top of it without asking permission. The MIT license is about as permissive as open-source gets. The only catch is you keep the license notice in any distribution.

Can I use Hermes Agent without self-hosting?

Yes — that's exactly what the Hermes Friend companion on Claw Friend is for. Select it during companion setup and you get the full Hermes Agent experience: persistent memory, self-improving skills, autonomous task execution, and 11+ platforms. No server, no Node.js, no API key management. Plans start at $9.99/mo.