Hermes-Agent
MIT-licensed, self-hosted AI agent with persistent memory that connects to 20+ chat platforms and builds its own reusable skills.
Hermes Agent is the right pick if you're technical and tired of re-explaining context to a bot that forgets you. Persistent memory, autonomous skill creation, and cross-session recall via FTS5 are the product — not a feature bullet — and the MIT license plus bring-your-own-provider (Nous Portal, OpenRouter, OpenAI, or any endpoint) means no lock-in. Against managed assistants like ChatGPT or Claude, you trade polish and zero setup for ownership, 20+ messaging surfaces, and memory that accumulates. Against IDE-bound coding agents like Cursor or Copilot, it's a different category: Hermes runs headless on a VPS and reaches you through Telegram, not inside your editor. The Graphana
Verified 1d ago · liveness 75/100 · cite: rightaichoice.com/tools/hermes-agent
- Developers who want a self-hosted, MIT-licensed agent that keeps memory and skills across sessions
- Power users automating recurring reports, backups, and briefings on a schedule
- Teams needing sandboxed subagent delegation without polluting main conversation context
- Privacy-conscious operators running local models on their own hardware with their own keys
- Users who want a plug-and-play consumer chatbot with no configuration
- Enterprises that need an SLA-backed hosted service with dedicated vendor support
- Non-technical users unwilling to touch a terminal or wire up messaging platform integrations
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip Hermes Agent if you want a zero-configuration consumer chatbot, need an IDE-embedded coding copilot, or require a vendor SLA — this agent expects you to run your own hardware and touch a terminal.
Nous Portal credit bonuses are capped by rollover limits — $10.00 on Plus, $50.00 on Super, $100.00 on Ultra — so unused monthly credits beyond the cap expire rather than accumulate.
Hermes Agent itself is free under the MIT license; the money goes to model access. Nous Portal credit bundles run $20/mo (Plus), $100/mo (Super), and $200/mo (Ultra), each with a 10% credit bonus. That sits well below managed enterprise agent platforms with per-seat contracts, and above simply pointing the agent at your own OpenRouter or OpenAI key. Solo developers usually start on Plus; heavy multi-bot teams push toward Super or Ultra.
In short
Hermes-Agent — MIT-licensed, self-hosted AI agent with persistent memory that connects to 20+ chat platforms and builds its own reusable skills. Best for Developers who want a self-hosted, MIT-licensed agent that keeps memory and skills across sessions, Power users automating recurring reports, backups, and briefings on a schedule, Teams needing sandboxed subagent delegation without polluting main conversation context. Free to start; paid plans from $20/mo.
What people actually say about Hermes-Agent — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
65 mentions across 6 sources (Hacker News, YouTube, Product Hunt, Stack Overflow, GitHub, Lemmy) · researched Jul 29, 2026.
Average across the 6 sources that answered — each source counts once, not each post.
- +Persistent memory across sessions via FTS5 and LLM summarization.
- +Autonomous skill creation that self-improves during use.
- +Supports 20+ messaging platforms for broad integration.
- +60+ built-in tools including web search, vision, and browser automation.
- +Runs on cheap VPS to GPU clusters, extremely flexible deployment.
- −Unintuitive and unresponsive interface frustrates new users.
- −Security concerns make it risky for handling sensitive data.
- −Hard to find genuinely useful standalone tasks for automation.
- −Website design described as eye-searing and off-putting.
- −Requires significant tinkering to set up and configure reliably.
- • You must pay for your own VPS or GPU compute; free tier offers no cloud hosting
- • Paid portal costs vary based on usage, not clearly advertised
Viability Score
How well maintained and how widely used is Hermes-Agent? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →Key Features
- Persistent memory across sessions for projects, conversations, and solved problems
- Auto-generates reusable skills from interactions and refines them during use
- Honcho dialectic user modeling that builds a deepening profile of you across sessions
- FTS5 cross-session recall with LLM summarization
- Natural-language scheduling via built-in cron with delivery to any platform
- Bot Mode: specialist bots with their own model, memory, skills, routines, and chats
- Isolated subagents with their own conversations, terminals, and Python RPC scripts
- Programmatic Tool Calling via execute_code for multi-step pipelines in single inference calls
- Web search, browser automation, vision, image generation, and text-to-speech
- Voice mode in CLI, Telegram, Discord, and Discord voice channels
- Seven terminal backends: local, Docker, SSH, Daytona, Singularity, Modal, Vercel Sandbox
- Container hardening and namespace isolation for sandboxed execution
- Command approval and authorization controls
- MCP integration for connecting external tool servers
- 20+ messaging platforms from one gateway including Telegram, Discord, Slack, WhatsApp, Signal, Matrix, Mattermost, Microsoft Teams, Google Chat, Email, SMS, DingTalk, Feishu, WeCom, Weixin, QQ Bot,
About Hermes-Agent
Hermes Agent is an open-source, MIT-licensed AI agent from Nous Research that you run on your own hardware — a $5 VPS, a GPU cluster, or serverless infrastructure like Daytona and Modal that hibernates when idle. Rather than one-off chat, it keeps a persistent memory of conversations, memories, and skills across sessions, so a task you taught it last month is still solved next time. The learning loop is the core claim: the agent auto-creates skills from your interactions, refines them during use, and maintains a deepening model of your preferences (Honcho dialectic user modeling) with FTS5 cross-session recall plus LLM summarization. You talk to it from CLI, desktop apps for macOS 12+ and Windows 10/11, or a messaging gateway covering Telegram, Discord, Slack, WhatsApp, Signal, Matrix, Mattermost, Microsoft Teams, Google Chat, Email, SMS, DingTalk, Feishu, WeCom, Weixin, QQ Bot, Yuanbao, BlueBubbles, and Home Assistant — 20+ platforms from one gateway, with the same memory behind all of them. Built-in cron scheduling runs reports, backups, and briefings unattended, and Bot Mode lets you assemble a durable team of specialist bots with their own model, memory, skills, routines, and chats that collaborate via group chats and @mentions. Delegation uses isolated subagents with their own conversations, terminals, and Python RPC scripts; Programmatic Tool Calling via execute_code collapses multi-step pipelines into single inference calls. Tooling spans 60+ built-in tools: web search, browser automation, vision, image generation, and text-to-speech. Execution is sandboxed across seven terminal backends — local, Docker, SSH, Daytona, Singularity, Modal, and Vercel Sandbox — with container hardening, namespace isolation, and command approval/authorization controls. MCP support connects external tool servers to the agent. A community Grafana observability project (Show HN, August 2026) adds monitoring and metrics. Research users get batch processing, trajectory export, and RL training with Atropos. Skills are portable via the agentskills.io open standard and the community Skills Hub. Model access is bring-your-own — Nous Portal, OpenRouter, OpenAI, or any compatible endpoint — and a single `hermes setup --portal` OAuth covers a model plus all four Tool Gateway tools. The agent itself is free; you pay only for whatever model provider or hosting you choose.
Behind the Verdict
The honest case for Hermes Agent rests on one thing the docs state plainly: it's the only agent with a built-in learning loop, and the architecture backs the claim. Agent-curated memory with periodic nudges, autonomous skill creation, self-improvement of those skills during use, FTS5 cross-session recall with LLM summarization, and Honcho dialectic user modeling form a genuine closed loop rather than a retrieval-augmented chat window. If you've ever wanted a bot to stop asking you the same setup questions, that is the mechanism. Strengths. Persistence is real and cross-surface: CLI, Telegram, Discord, Slack, WhatsApp, Signal, Matrix, Mattermost, Teams, Google Chat, Email, SMS, DingTalk, Feishu, WeCom, Weixin, QQ Bot, Yuanbao, BlueBubbles, and Home Assistant all route through one gateway to one memory. Voice mode works in CLI, Telegram, Discord, and Discord voice channels. Bot Mode is more useful than it first appears — specialist bots with separate models, memories, skills, routines, and chats that coordinate via @mentions is how you'd actually split a personal-assistant workload. Programmatic Tool Calling via execute_code collapsing multi-step pipelines into single inference calls is a cost and latency win you notice at volume. Seven sandbox backends (local, Docker, SSH, Daytona, Singularity, Modal, Vercel Sandbox) with container hardening and namespace isolation is more than most self-hosted agents offer; Daytona and Modal give serverless persistence so an idle agent costs nearly nothing. MCP support and the agentskills.io-compatible Skills Hub mean you aren't locked into the built-in 60+ tools. Research users get batch processing, trajectory export, and RL training with Atropos — rare in this category. Where it doesn't fit. Setup assumes competence: the installer is a curl script or a desktop app, and connecting messaging platforms is documented but manual. There's no SLA, no vendor accountability for your uptime, and no consumer-grade onboarding. It is explicitly not a coding copilot in an IDE, and not a plug-and-play chatbot. And a hard operational truth from the vendor's own FAQ: stopping a local agent ends its running process, so scheduled tasks that need an active agent won't fire while it's down — hosted agents are managed separately in Nous Portal. Back up your Hermes data directory when you move machines. On pricing: Hermes Agent itself is free under the MIT license, and model providers or optional cloud hosting carry their own costs. Nous Portal sells credit bundles — Plus at $20/mo, Super at $100/mo, and Ultra at $200/mo, each including a 10% bonus in monthly credits ($22.00, $110.00, $220.00 respectively) with rollover caps of $10.00, $50.00, and $100.00 — alongside a $0 Free tier limited to free models and standard rate limits. The bonus and rollover caps are the detail most reviewers miss: credits are use-it-or-lose-it beyond the cap. The call: install it if you want ownership, cross-platform reach, and memory that
Researching Hermes-Agent? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas Hermes-Agent actually fits — and what changes day-one when you adopt it.
Install via the curl script on Linux, run `hermes setup --portal` so one OAuth covers a model plus the four Tool Gateway tools, then connect a Telegram bot token and set a nightly cron job that summarizes an inbox and posts it to Telegram.
Outcome: A persistent agent that answers from Telegram, remembers your project context, and delivers a briefing every morning without you logging in.
Connect Slack and email through the messaging gateway, write natural-language schedule entries for a weekly report and a nightly backup check, and let the built-in cron deliver results to Slack.
Outcome: Recurring reports and backup verification land in Slack unattended, and the agent refines the skill each time it runs.
Use Docker or Modal as the sandbox backend for isolation, spawn isolated subagents for parallel workstreams, then export trajectories and run RL training with Atropos.
Outcome: Parallel, sandboxed agent runs with trajectory data ready for training, without polluting the main conversation context.
Use Cases
- Automate daily report generation and briefings via built-in cron delivered to any chat platform
- Delegate web research and data extraction to isolated subagents so your main context stays clean
- Run a personal assistant that remembers project preferences and answers you on Telegram, Discord, or Slack
- Assemble a team of specialist Bots that coordinate through group chats and @mentions
- Control Home Assistant smart-home devices by voice or text through the same agent
- Collapse multi-step pipelines into single inference calls with execute_code programmatic tool calling
- Export trajectories and run RL training with Atropos for agent research
- Install a skill from the community Skills Hub or author your own via the agentskills.io standard
Models Under the Hood
as of 2026-09-29
Limitations
- Hermes Agent is free and open source under the MIT license, but model providers and optional hosted services have their own pricing, so you must supply your own provider or connect to Nous Portal for model and tool credits.
- Cloud hosting is optional yet requires separate configuration for always-on deployments.
- Setup is done through a desktop installer or terminal commands.
- Stopping a local agent ends its running process — saved conversations, memories, and skills persist, but tasks that need an active agent will not run while it is stopped, and hosted agents are managed separately in Nous Portal.
- Back up your Hermes data directory when moving to another computer.
as of 2026-10-07
Verification history
We have re-verified Hermes-Agent 8 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Showing the 6 most recent of 8 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published Hermes-Agent tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Free
$0/mo
Ideal for
Developers who already pay for a model provider and just want the MIT-licensed agent plus messaging integrations at no cost
What this tier adds
Starting tier — $0 with free models only, standard rate limits, and no monthly credits
Plus
$20/mo
Ideal for
Solo developer or power user who wants hosted model and tool access without managing separate provider keys
What this tier adds
Adds $22.00 monthly credits with a 10% bonus, access to 200+ models, hosted tool usage, high rate limits, and a $10.00 rollover cap
Super
$100/mo
Ideal for
Heavy individual user or small team running multiple specialist Bots and frequent scheduled automations
What this tier adds
Raises monthly credits to $110.00 with a 10% bonus and lifts the rollover cap to $50.00 over Plus
Ultra
$200/mo
Ideal for
Research groups or operations teams pushing large multi-bot and subagent workloads continuously
What this tier adds
Top tier at $220.00 monthly credits with a 10% bonus and a $100.00 rollover cap, the highest credit ceiling offered
Where the pricing makes sense
The company stage and team size where Hermes-Agent's pricing actually pencils out — and where peers do it cheaper.
Hermes Agent itself is free under the MIT license; the money goes to model access. Nous Portal credit bundles run $20/mo (Plus), $100/mo (Super), and $200/mo (Ultra), each with a 10% credit bonus. That sits well below managed enterprise agent platforms with per-seat contracts, and above simply pointing the agent at your own OpenRouter or OpenAI key. Solo developers usually start on Plus; heavy multi-bot teams push toward Super or Ultra.
Setup time & first value
How long it actually takes to get something useful out of Hermes-Agent — broken out by persona, not the marketing-page minute.
On Linux, macOS, or WSL2 the curl installer gets you to a working agent in about 60 seconds, and `hermes setup --portal` adds a model plus the four Tool Gateway tools in one OAuth step. The desktop app for macOS 12+ or Windows 10/11 is the fastest route for non-terminal users. Budget 15-30 minutes if you're wiring several messaging platforms, and longer for an always-on hosted deployment.
Switching to or from Hermes-Agent
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From a single-platform chat bot: connect the platform you already use as one of the 20+ gateway integrations and point Hermes at your existing model provider keys.
- →From ChatGPT or Claude for recurring tasks: move scheduled work to Hermes cron jobs and let persistent memory replace re-pasting context each session.
- →From a homegrown script collection: port them as skills under the agentskills.io standard and let Hermes refine them during use.
- →From an IDE-bound coding agent: keep the IDE tool for editing and route research, scheduling, and messaging work to Hermes on a VPS.
- ↗To a managed consumer assistant: export your Hermes data directory as a record, since skills and memories don't port to closed platforms.
- ↗To a hosted agent platform with an SLA: rebuild scheduled routines as platform-native workflows and re-connect messaging channels on the new service.
- ↗To another open-source agent: skills written to the agentskills.io open standard are portable, so carry those across and rebuild memory from your exported conversations.
- ↗To a coding copilot in your IDE: keep Hermes for scheduling and messaging alongside it rather than replacing one with the other.
Integrations
Resources & Guides
Tutorials & Learning
Official links
Tools that pair well with Hermes-Agent
Common stack mates teams adopt alongside Hermes-Agent, with the specific reason each pairing earns its keep.
Featured Head-to-Head Comparisons
Hermes Agent vs Presto Voice
Presto Voice is purpose-built for QSR drive-thru automation with proven ROI, while Hermes-Agent is a general-purpose open-source agent for developers. Choose Presto if you run a drive-thru chain and need a turnkey voice AI solution; choose Hermes if you want a customizable, self-hosted autonomous agent that can learn and automate across platforms.
Hermes Agent vs Locus Robotics
If you run a high-volume warehouse needing proven 2-3x productivity gains with minimal disruption, Locus Robotics's RaaS model and new Locus Array Physical AI are unmatched. But if you're a developer or power user wanting a self-hosted, autonomous AI agent that learns and works across messaging platforms, Hermes-Agent is a revolutionary free tool. These tools serve completely different worlds — choose based on whether you move physical boxes or digital tasks.
Hermes Agent vs Truleo
Truleo and Hermes-Agent serve entirely different audiences. Truleo is a specialized law enforcement intelligence platform, connecting siloed data to generate leads and speed up report writing. Hermes-Agent is a general-purpose open-source agent for automating workflows across messaging platforms. Choose Truleo for police department use; choose Hermes-Agent for personal or team automation needs. They are not direct competitors.
Alternatives to Hermes-Agent
View all
Popular in Agent Frameworks & Orchestration
Temporal AI
Temporal is the durable execution platform that keeps AI agents and long-running workflows alive through crashes, retries, and abandoned
Frequently Asked Questions
Best-of guides
Used Hermes-Agent? Help shape our editorial sentiment research.