Self-hosted personal agents
Run one persistent assistant on a local machine, VPS, container, remote server, or serverless sandbox.
Independent tool overview
Hermes Agent at a glance
Hermes Agent is Nous Research's active MIT-licensed personal agent for terminal, desktop, messaging, scheduled automation, persistent memory, learned skills, and delegated tool use. It offers unusual model and hosting flexibility, but it is an operator-controlled agent with access to files, commands, credentials, and external services, so secure configuration matters as much as model quality.
Visit the official Hermes Agent site ↗
- Developer
- Nous Research
- License
- MIT
- Interfaces
- Desktop, terminal, messaging, gateway, Python
- Model choice
- Nous Portal, external APIs, or self-hosted endpoints
- Core differentiator
- Persistent memory and self-improving skills
Overview
What Hermes Agent is
Hermes Agent can run as a desktop app, terminal interface, local or remote gateway, Python library, or messaging bot. A single installation can be reached from services including Telegram, Discord, Slack, WhatsApp, Signal, email, and Home Assistant.
Its defining feature is persistence. Hermes stores session history, maintains curated memory and user-profile files, searches past conversations, and can create or refine reusable skills after complex work. Scheduled jobs and subagents let it continue beyond a single chat turn.
The project is model-agnostic. Users can connect Nous Portal, OpenRouter, OpenAI, Anthropic-compatible endpoints, or self-hosted models and can switch models without changing the agent code. Terminal backends include local execution, Docker, SSH, Singularity, Modal, and Daytona.
Hermes includes meaningful security controls—messaging allowlists, pairing, approval modes, deny rules, protected credential paths, container isolation, prompt-injection scanning, and session separation. Those controls still require careful setup; broad permissions, local execution, unattended jobs, or YOLO mode can create serious risk.
Use cases
Who Hermes Agent is best for
The strongest fit depends on the job you need the product to complete, not the size of its feature list.
Cross-platform automation
Reach the same agent from messaging apps and schedule reports, backups, briefings, audits, and other recurring work.
Technical users who want model choice
Combine an open agent runtime with different hosted or local models instead of committing to a single model provider.
Capabilities
Core Hermes Agent features
Persistent memory
Maintains curated agent memory, a user profile, searchable sessions, and optional external memory-provider integrations across conversations.
Reusable skills
Can create procedural skills from successful work and improve those skills as it uses them again.
Messaging gateway
Connects the same agent to Telegram, Discord, Slack, WhatsApp, Signal, email, Home Assistant, and other supported channels.
Scheduled automation
Natural-language cron jobs can run unattended and deliver results back to configured platforms.
Subagent delegation
Spawns isolated workers with separate conversations and terminals for parallel tasks or scripted tool pipelines.
Flexible execution backends
Runs commands locally or inside Docker, SSH, Singularity, Modal, Daytona, and other documented sandbox backends.
Broad tool support
Includes terminal, browser, search, vision, image generation, speech, MCP, and other toolsets that can be enabled or restricted.
Process
How the Hermes Agent workflow works
Step 1
Choose the safest runtime
Prefer a dedicated user, isolated container or remote sandbox, narrow workspace, and egress controls rather than broad access to a personal machine.
Step 2
Connect one model first
Configure Nous Portal or another provider, cap iterations and spending, and validate basic chat and tool behavior before adding more services.
Step 3
Lock down messaging
Use pairing or explicit user allowlists, keep allow-all settings off, protect bot tokens, and test that unauthorized users are denied.
Step 4
Enable tools gradually
Use smart or manual approvals, user-defined deny rules, safe write roots, restricted MCP credentials, and human confirmation for consequential actions.
Step 5
Inspect memory and schedules
Review stored sessions, memory, skills, cron jobs, logs, tool permissions, costs, and permanent approvals on a recurring basis.
Cost
Hermes Agent pricing and free plan
Hermes Agent's software is free under the MIT license. Total cost depends on the model provider, hosting, storage, search or media tools, messaging services, and how many tool iterations an agent runs. Nous Portal currently offers a free tier plus $20, $100, and $200 monthly plans with bonus usage credits; users can instead bring another API or self-host a model. A small VPS can be inexpensive, while frontier models and long autonomous runs can cost far more than the server.
Hermes Agent software
$0
Open-source agent runtime under the MIT license.
- Desktop and terminal options
- Model usage and hosting are separate
- You operate updates, security, backups, and availability
Nous Portal Free
$0/month
Access to free models at standard limits.
- Free models only
- No monthly paid credits
- Can be used with Hermes Agent
Nous Portal Plus
$20/month
Entry paid model and hosted-tool bundle.
- $22 monthly credits
- $10 rollover cap
- 200+ models and high rate limits
Nous Portal Super
$100/month
Larger monthly model and tool allowance.
- $110 monthly credits
- $50 rollover cap
- 200+ models and high rate limits
Nous Portal Ultra
$200/month
Highest listed individual Portal allowance.
- $220 monthly credits
- $100 rollover cap
- 200+ models and high rate limits
Bring your own provider or hosting
Varies
Use OpenRouter, direct model APIs, a local endpoint, or infrastructure you operate.
- Provider token pricing applies
- VPS, GPU, storage, networking, backup, and tool costs may apply
- Set iteration and budget limits to avoid runaway usage
Pricing checked . Check current pricing at the source ↗
Assessment
Hermes Agent strengths and limitations
Where it stands out
- MIT-licensed runtime with no required model-provider lock-in
- Persistent memory, session search, learned skills, schedules, and subagents support work beyond a single conversation
- Desktop, terminal, messaging, gateway, and library interfaces cover many operating styles
- Local, remote, container, and serverless execution backends provide unusual deployment flexibility
- Documentation includes concrete authorization, approval, sandboxing, credential, and prompt-injection controls
What to consider
- Setup and secure operation require more technical judgment than a fully managed consumer assistant
- Agent memory is curated and bounded; saved history does not mean every prior detail is automatically present or correct in a new turn
- Model calls, long contexts, subagents, browser work, and high iteration limits can create substantial and unpredictable usage costs
- Local or SSH execution can expose files, credentials, processes, and networks if permissions are too broad
- Write guards reduce accidental damage but the terminal runs with the configured operating-system user's authority unless it is isolated
- YOLO mode and unattended cron approval can bypass important human checks and should be restricted to disposable or tightly controlled environments
Compare
Hermes Agent alternatives
The right alternative depends on the specific output, workflow, controls and budget your project requires.
Agents
OpenClaw
Consider OpenClaw for another self-hosted messaging-first personal agent with broad action capabilities.
Explore OpenClaw →Coding
Claude Code
Consider Claude Code when the primary need is a managed-model terminal coding agent rather than a persistent cross-platform personal agent.
Explore Claude Code →Coding
Codex App
Consider the Codex app for a desktop coding workflow with task isolation, repository tooling, and OpenAI-managed models.
Explore Codex App →Questions
Hermes Agent FAQs
Is Hermes Agent free?
The Hermes Agent software is free under the MIT license. You still pay for any hosted models, servers, storage, messaging services, or external tools you choose to use.
Which models can Hermes Agent use?
Hermes supports Nous Portal, OpenRouter, direct provider APIs, compatible custom endpoints, and self-hosted models. The active model can be changed without rewriting the agent.
Does Hermes remember previous conversations?
Yes. It stores sessions, supports full-text search, and maintains curated memory and user-profile files. Memory is selective and bounded, so important facts should still be verified.
Can Hermes run through Telegram or Discord?
Yes. Its gateway supports Telegram, Discord, Slack, WhatsApp, Signal, email, Home Assistant, and other channels. Configure pairing or allowlists before exposing a bot.
Is Hermes Agent safe to run autonomously?
It includes strong configurable controls, but no general-purpose tool-using agent is automatically safe. Use isolation, narrow permissions, deny rules, approvals, credential filtering, budget caps, backups, and human review for high-impact actions.
Bottom line
Our Hermes Agent verdict
Hermes Agent is a strong option for technical users who want an open, persistent agent they can reach from many surfaces and pair with almost any model. Memory, skills, schedules, messaging, and subagent delegation make it more than a coding chat wrapper. Its flexibility also transfers operational responsibility to the user: the best deployment is isolated, narrowly authorized, cost-capped, backed up, and regularly audited.
Visit Hermes Agent website ↗Similar Agents Tools
View all Agents AI tools →
Perplexity Computer
Perplexity Computer - Perplexity's multi-model agent system orchestrating 19 AI models for long-running tasks
Luma Agents
Luma Agents
.jpeg)
Notion Custom Agents
Notion Custom Agents - Always-on AI agents for automating team workflows across tools
Codex Security
Codex Security - OpenAI's new security agent for scanning repos and patching code vulnerabilities
AI training for the future of work.
Get access to all our AI courses, hundreds of real-world AI use cases, live expert-led workshops, an exclusive network of AI early adopters, and more.
AI Courses
Get unlimited access to all of our current & upcoming industry-specific AI courses for the duration of your subscription.
Daily Guides
To keep up with the rapid pace of AI, our team publishes AI implementation guides daily. Our library contains 300+ practical use cases to automate real-world work.
Workshops
Join weekly, live, interactive sessions with industry leaders who are at the forefront of AI for hands-on implementation guidance and exclusive insights.
Community
Network with an exclusive community of AI-first professionals who are working smarter with AI. Learn how early adopters are using AI in their work and businesses.