Source-backed analysis12
Review basis
Research-based review: the AppDuo Editorial Team compiled it from official pages, documentation, app stores, press and user feedback, without hands-on testing. Every claim links to a source with the date it was checked.
12 reference sourcesInformation updated Nov 11, 2026
How we review
Hermes Agent Review
Nous Research's open-source AI agent with persistent memory and self-written skills
By AppDuo Editorial Team · Fact-checked
- macOS
- Windows
- Linux
On this page
Decide at a glance
Our verdict
Hermes Agent suits technical users who want a self-hosted, open-source AI agent that remembers projects and answers through Telegram, Discord or Slack. In exchange you set up and pay for models yourself, and the vulnerabilities published in 2026 are worth reviewing before letting an agent run commands on your machine.
Best for: Developers and technical users who want a self-hosted agent under the MIT license
Skip it if: Users who only want a simple AI chat app and do not want to configure models or runtimes
Pros
- Open source under the MIT license; the GitHub repository shows about 262.6k stars and 58.8k forks (Nov 2026)
- Memory persists across sessions, with generated skills, natural-language scheduling and isolated subagents
- Wide channel support: Telegram, Discord, Slack, WhatsApp, Signal, email, Teams, Google Chat and more
Cons
- An independent audit published on 11 Apr 2026 reported 4 Critical and 9 High findings in v0.8.0; several CVEs were also published during 2026
- Model costs are separate: API keys always draw on prepaid credits, and Nous Portal plans carry daily, weekly and monthly limits the pricing page does not quantify
All pros and cons · Who is it for?
More pros
- You choose where it runs (local, Docker, SSH, Modal or a cloud agent) and which model it uses (Nous Portal, OpenRouter, OpenAI or any endpoint)
More cons
- Under the Nous Portal policy, data may be used to improve services and shared with AI-training providers unless you turn on Privacy Mode
- Few public ratings: 16 on the Microsoft Store (Nov 2026), none found on the App Store, Google Play or major review sites
Best for
- People who want one agent with one memory across several chat channels and scheduled tasks
Skip it if
- Organisations needing security certifications or support SLAs: none were found in the pages read
Experience & key features
What is Hermes Agent?
Hermes Agent is an open-source AI agent from Nous Research, released under the MIT license. You install it with a one-line terminal command on macOS, Linux or Windows, or download the desktop app for macOS 12+ and Windows 10/11 (the Windows app is also on the Microsoft Store). The homepage describes one agent with one memory shared across the CLI, Telegram, Discord, Slack, WhatsApp, Signal and email.
How it works and key features
Per the homepage and docs, conversations, memories and skills persist between sessions, and stopping a local agent does not delete them. The agent can schedule tasks in natural language, delegate to isolated subagents, search the web, drive a browser, generate images and read text aloud. Commands can run locally, in Docker, over SSH, on Modal and on other backends. For models, the docs list Nous Portal, OpenRouter, OpenAI or any endpoint.
Pricing and value
The self-hosted software is free; you pay only the model provider you choose. Nous Portal has four monthly plans: Free at $0 (free models only), Plus at $20, Super at $100 and Ultra at $200. The three paid plans include capped Hermes Agent usage and one cloud agent; the pricing page does not state the included credit amounts. API keys always draw on credits, topped up in $10, $20, $50, $100 or $200 amounts via Stripe. There is no yearly billing, and the pricing page states no refund terms.
More analysis
Who it's for and what to weigh
Hermes Agent suits technically capable users who want control over where the agent runs and which model it uses. OpenClaw is a comparable open-source agent; Claude, ChatGPT and Manus are hosted services that need less setup. Because the agent can run commands and read files, review the published vulnerabilities, keep it updated and use the command approval and container isolation the docs mention.
Screenshots
Plans & pricing
Prices last verified Nov 11, 2026
Hermes Agent (open source)
PersonalFree
Free, open-source software under the MIT license. You pay separately for AI model usage (Nous Portal, OpenRouter, OpenAI or another endpoint) and for any cloud hosting.
Nous Portal Free
PersonalFree
Free models only, standard rate limits, no monthly credits.
Nous Portal Plus
Personal$20/mo
Includes Hermes Agent usage within limits that reset daily, weekly and monthly (pricier models use it up faster), plus one cloud agent with up to 4 GB RAM and 25 GB disk. Bigger sizes and extra agents use credits.
Nous Portal Super
Personal$100/mo
Same usage terms as Plus; cloud agent with up to 4 GB RAM and 100 GB disk.
Nous Portal Ultra
Personal$200/mo
Same usage terms as Plus; cloud agent with up to 8 GB RAM and 100 GB disk.
- Free plan limits
- The self-hosted software is fully free; Nous Portal Free is limited to free models.
- Free trial
- No free trial
Cancellation & refunds
The Nous Portal pricing page states no refund terms and only links to its Terms.
Privacy
Per the homepage, a local install keeps conversations, memory and skills in a Hermes data directory on your machine. For Nous Portal and cloud agents, the privacy policy (updated 10 Nov 2026) lists prompts, outputs, uploads, API calls, usage telemetry, device, browser and billing data among what Nous collects. Nous may use this data to improve its services, share it with providers that support AI training, and use aggregated or de-identified data for AI training. When a feature relies on a third-party model, prompts and outputs go to that provider. Privacy Mode (Account Settings > Privacy Settings) stops Nous storing and using inference payloads for training, product improvement and support; it applies only from when it is turned on, excludes operational metadata and does not change third-party processing. Account deletion takes effect 7 days after the request. Published vulnerabilities are listed under Controversies below; the project's detailed security documentation could not be read for this review.
Hermes Agent alternatives
7.5/10ChatGPT
OpenAI's flagship AI assistant for writing, research, coding, images and voice
7.4/10Claude
Anthropic's AI assistant, strong for coding and writing, with tight usage limits
Controversies & incidents
Issues documented by the press, regulators, courts or the vendor itself. Every item is sourced; unproven claims are labelled as allegations. Last researched: Nov 11, 2026
- Low impactAllegedSecurity ·
Reports of attackers using Hermes Agent in intrusions
The AI Incident Database lists two July 2026 intrusions in which attackers reportedly used Hermes Agent, in one case in unattended YOLO mode. The reports describe misuse of the tool, not a flaw in Hermes.
Read the full account
The AI Incident Database lists Hermes Agent as an implicated system in two incidents. In the first (1–4 Jul 2026), suspected China-linked hackers reportedly used a multi-agent framework built on Hermes and OpenClaw to breach Taiwanese government systems, compromising 85 credentials and exfiltrating more than 2,564 personnel records; Taiwan later confirmed an overseas AI-assisted campaign but did not attribute it to China. In the second, Hunt.io reported that an unidentified actor ran Hermes in unattended YOLO mode on Thailand's Ministry of Finance network; researchers found no exfiltration and the ministry had not confirmed a breach. Separately, Borncity reported (3 Aug 2026), without naming its source, that an actor pairing Hermes with DeepSeek compromised more than 460 internet-facing systems. None of these reports alleges a vulnerability in Hermes or wrongdoing by Nous Research.
Sources:AI Incident Database: Hermes Agent entityBorncity: KI-Agent-Sicherheit, 460 Systeme durch Hermes-Exploits kompromittiert (2026-08-03)
- Medium impactResolvedSecurity ·
File-permission flaw exposed conversation history to other local users (CVE-2026-53870)
Versions before 0.16.0 created two data files readable by every local user. The flaw is rated Medium and fixed in 0.16.0.
Read the full account
According to the CVE record summarised by OffSeq (published 17 Jun 2026, assigned by VulnCheck), versions before 0.16.0 created response_store.db and webhook_subscriptions.json with mode 0o644, so low-privileged local users could read conversation history, tool payloads, prompts and per-route HMAC secrets. The CVSS 4.0 score is 6.8 (Medium), with no known exploitation. The recommended action is to upgrade to 0.16.0.
Vendor response
Nous Research shipped version 0.16.0 containing the fix, per the CVE record.
- Medium impactAllegedSecurity ·
Independent audit alleges serious flaws in v0.8.0, plus three early-2026 CVEs
An independent researcher reported 4 Critical and 9 High findings in v0.8.0; a Cloud Security Alliance note also lists 3 Medium CVEs.
Read the full account
Per a Cloud Security Alliance (CSA) research note (4 May 2026), an audit that independent researcher @Anic888 posted on 11 Apr 2026 listed Critical findings in the default configuration: a terminal tool passing commands to a shell guarded only by bypassable regex checks, a file-read tool with no deny list that can reach SSH keys, container environments skipping all approval checks, and an agent able to write skills that persist across sessions. High findings included a YOLO mode that disables checks, prompt-injectable LLM auto-approval, an opt-in write sandbox and unsandboxed plugins. These findings have no CVE IDs and CSA gives no patch status for them. The three Medium CVEs are CVE-2026-7396 (path traversal in the WeChat Work adapter, no patch cited by CSA), CVE-2026-7397 (symlink following in file tools, fixed in v0.9.0) and CVE-2026-6829 (path traversal in hermes-webui, fixed from 0.50.34). CSA documents no response from Nous Research and notes only that the project has since reached v0.12.0.
Sources:Cloud Security Alliance research note on Hermes Agent CVEs (2026-05-04)
Frequently asked questions
Is Hermes Agent free?
The software is free and open source under the MIT license. You still pay the chosen provider for AI model usage; Nous Portal has a Free tier limited to free models.
Which chat channels does Hermes Agent support?
The docs list the CLI, Telegram, Discord, Slack, WhatsApp, Signal, Matrix, Mattermost, email, SMS, DingTalk, Feishu, WeCom, Microsoft Teams, Google Chat and more.
Does Hermes Agent support Vietnamese?
Only English and Simplified Chinese documentation could be verified. Vietnamese output depends on the AI model you choose; Nous has no official Vietnamese-specific statement.
Is my data used for training?
With Nous Portal, the policy says Nous may use data to improve services, share it with AI-training providers and train on aggregated or de-identified data, unless you turn on Privacy Mode (which applies only from then on). When running locally, data stays in a Hermes directory on your machine, though prompts still go to the model provider you choose.
Review evidence & sources12
Scores
- Ease of use
- 6.2
- Features
- 8.2
- Value for money
- 8.0
- Performance
- 6.8
- Support
- 6.0
- Privacy & security
- 6.5
Why this score
A one-line install and a desktop app for macOS 12+ and Windows 10/11 (the Windows app is on the Microsoft Store), but you still pick a model provider, chat channels and a runtime. Docs are in English and Simplified Chinese; no other interface languages were verified.
Why this score
Persistent memory, generated skills, scheduling, subagents, web search, browser automation, image generation, text-to-speech, several sandbox backends and a long list of chat channels. Broad among open-source agents, but no independent measurements of feature quality were read.
Why this score
Free MIT-licensed software. Nous Portal offers Free, Plus at $20, Super at $100 and Ultra at $200 a month. Marked down because costs are hard to predict: pricier models use up allowances faster, the pricing page does not state included credit amounts, and bigger sizes or extra agents draw on credits.
Why this score
No independent benchmarks were found, and results depend heavily on the model you choose. The README warns that Windows Defender or Bitdefender may quarantine the bundled uv.exe, which it calls a false positive. The score is conservative because measurements are missing.
Why this score
Docs, Discord, GitHub Issues and a Skills Hub are available. No staffed support contact or response-time commitment was found, and the Nous Portal pricing page states no refund terms.
Why this score
Credit for local running that keeps data on your machine and for Privacy Mode. Marked down because by default the policy allows data use for service improvement and sharing with AI-training providers, an independent audit reported serious findings in v0.8.0, and a file-permission flaw (CVE-2026-53870) affected versions before 0.16.0.
What users say
- Microsoft Store (US)4.7/5 · 16
Sources
- Vendor-providedOfficial Hermes Agent homepageNov 11, 2026
- Vendor-providedNous Portal pricingNov 11, 2026
- Vendor-providedHermes Agent documentationNov 11, 2026
- Vendor-providedNous Portal privacy policyNov 11, 2026
- Vendor-providedNous Research homepageNov 11, 2026
- Vendor-providedGitHub: NousResearch/hermes-agentNov 11, 2026
- pressCloud Security Alliance research note on Hermes Agent CVEs (2026-05-04)Nov 11, 2026
- Community-reportedOffSeq Threat Radar: CVE-2026-53870Nov 11, 2026
- pressFortune: Paradigm backs Nous Research (2025-04-25)Nov 11, 2026
- storeMicrosoft Store: Hermes AgentNov 11, 2026
- Community-reportedAI Incident Database: Hermes Agent entityNov 11, 2026
- pressBorncity: Hermes misuse report (2026-08-03)Nov 11, 2026
Specs & compatibility
- Developer
- Nous Research
- Headquarters
- United States
- Launched
- 2026
- Category
- AI
Capabilities
- AI features
- Open source
- Self-hostable
- Integrations
Platforms
- macOS
- Windows
- Linux
Interface languages
- English
Integrations
- Telegram
- Discord
- Slack
- Signal
- Microsoft Teams
- Google Chat
- Home Assistant
- OpenRouter
- OpenAI
- Docker
- Modal
Support
- Help center
- Community forum
Change history
- Update.The project's GitHub repository shows about 262.6k stars and 58.8k forks at the time of AppDuo's check.
- Update.CVE-2026-53870 (overly permissive file modes) is published; fixed in version 0.16.0.
- Update.An independent researcher posts a security audit of v0.8.0 to the project's public issue tracker, reporting 4 Critical and 9 High findings.
Full change history (4)
- Company.Per Fortune, Nous Research raises a $50 million Series A financed almost entirely by Paradigm; an anonymous source puts the token valuation at $1 billion.
About the author
Reviews & comments
Written by readers with a verified email address.
No reviews yet. Have you used Hermes Agent? Share your real experience.
Reader reviews are counted separately and do not affect the Editorial Team’s score.
No comments yet. Be the first to share your experience.
Comments are readers’ opinions, not content from the AppDuo Editorial Team.