Cadenya
Paid
Developers building agentic workflows
Hermes Agent
FreeOpen SourceAPISelf-HostedAgentic
Hermes Agent, from Nous Research, is an open-source agent that remembers things between sessions and saves reusable skills. You can reach it from a CLI or through messaging apps like Telegram and Discord.
Free (MIT, self-hosted) · Free tier: The software is free under the MIT license. You pay for your own server and for whichever model provider you connect. · Checked Jul 1, 2026
Similar tools
Claude CodePaidAnthropic's coding agent for editing repositories, running commands and tests, and reviewing…
WorkBuddyFreemiumWorkBuddy runs as a local-first agent on the desktop, autonomously chaining file access, web…
Hermes DesktopFreeHermes Studio is an open-source, self-hosted dashboard that wraps Hermes Agent in a control…
ChatGPTFreemiumOpenAI's assistant for writing, research, analysis and file creation through Chat, Work and…
Get This Tool
License: MIT Any use incl. commercial
Local-run terms: MIT license permits unrestricted use, modification, and redistribution for any purpose, including commercial, provided the license and copyright notice are included. No restrictions on local deployment, data ownership, or commercial applications.
Screenshots 4
Summary
We ran Hermes Agent v0.22.1 on the same three jobs as Goose and OpenCode: create a file, summarize a report, and count CSV rows with a shell command, three runs each. With a small local model (qwen2.5:3b) it passed none of the 9. That isn't really a verdict on Hermes: the model kept calling Hermes' tool dispatcher instead of the file and shell tools, and Hermes' docs ask for a far bigger model and a 64K context, while our setup gave it 32,768 tokens. With Claude Opus 5.5 it passed all 9. It was also the heaviest of the three agents: a fixed prompt of about 11,500 tokens on every model call, about 47,100 tokens per task, $0.27 for all nine runs, and the slowest time on every task. With the local model it tried to reach GitHub, OpenRouter and Nous Research hosts on every run; with Claude it contacted only Anthropic's API.
Bottom line: Pick Hermes if you want memory, saved skills and messaging access, and you'll pair it with a large model. It's a poor fit for small local models, and it uses far more tokens per task than Goose or OpenCode.
Pricing Plans
FreeLast verified 4 months ago- Price
- Free (MIT, self-hosted)
- Free Tier
- The software is free under the MIT license. You pay for your own server and for whichever model provider you connect.
Self-hosted Hermes Agent
Free
Open source, MIT licensed. You pay only for your own server and for whichever model provider you connect.
- MIT licensed open source
- Maximum control over stack and data
- Custom infrastructure options
- Private deployments
- Deep customization
View full pricing on hermes-agent.nousresearch.com →
Pricing may have changed since last verified. Check the official site for current plans.
We installed and ran it
- How we installed it
- Vendor one-line install script (browser and computer-use extras skipped)
- Install script run time
- 151.5 s
- Docker image size on disk
- 4.17 GB
- Local model we used
- qwen2.5:3b via Ollama
- Agent tasks passed with the local model
- 0 of 9
- Fixed prompt sent with every model call
- ~11,500 tokens
- Outside hosts it tried to reach (we blocked them)
- api.github.com, hermes-agent.nousresearch.com, openrouter.ai, portal.nousresearch.com, raw.githubusercontent.com
| Task (qwen2.5:3b) | Passed | Median time |
|---|---|---|
| Create a file with given text | 0 / 3 | 177.4 s |
| Read a report and write a 2-sentence summary | 0 / 3 | 156.1 s |
| Count the data rows in a CSV with a shell command | 0 / 3 | 100 s |
With Claude Opus 5.5 (Anthropic API)
- Model
- Claude Opus 5.5 (claude-opus-5-5)
- Agent tasks passed
- 9 of 9
- API cost for all 9 runs (list price)
- $0.27
- Tokens per task (input + output)
- ~47,100
- Tokens per model call
- ~17,700
- Model calls, all 9 runs
- ~24
- Input tokens read from the prompt cache
- 94%
- Other hosts it tried to reach (the proxy refused them)
- None
| Task (Claude Opus 5.5) | Passed | Median time |
|---|---|---|
| Create a file with given text | 3 / 3 | 18.8 s |
| Read a report and write a 2-sentence summary | 3 / 3 | 24.8 s |
| Count the data rows in a CSV with a shell command | 3 / 3 | 22.3 s |
Same three tasks, Claude Opus 5.5 driving each agent
| Measured | Hermes Agent | Goose | OpenCode |
|---|---|---|---|
| Tasks passed with Claude Opus 5.5 (of 9) | 9 | 9 | 9 |
| Median time: Create a file with given text (s) | 18.8 | 8.1 | 11 |
| Median time: Read a report and write a 2-sentence summary (s) | 24.8 | 10.3 | 16.3 |
| Median time: Count the data rows in a CSV with a shell command (s) | 22.3 | 11.6 | 10.9 |
| Model calls, all 9 runs | 24 | 26 | 25 |
| Tokens per task, input + output (approx.) | 47,100 | 7,800 | 29,200 |
| API cost, all 9 runs (US$, list price) | 0.27 | 0.14 | 0.19 |
| Outside hosts it tried to reach besides the API (blocked) | 0 | 1 | 2 |
| For reference: tasks passed with qwen2.5:3b on 3 CPU cores (of 9) | 0 | 5 | 3 |
All three ran on October 16, 2026, one after another, with Claude Opus 5.5 (claude-opus-5-5) through Anthropic's API and each agent's default settings: three runs per task, each in a fresh container, graded on the files left behind. The containers had no internet; a proxy let them reach api.anthropic.com and nothing else. Times include the round trips to Anthropic's API, so they show the agent and the API together, not this machine's speed. Tokens come from each agent's own output (Hermes Agent's model-call count is approximate); cost is those tokens at Anthropic's list price. The qwen2.5:3b row is from the earlier local-model runs (Goose and Hermes Agent on October 15, OpenCode on October 16).
What we noticed
- With a 3B model, Hermes completed none of the 9 tasks: the model called Hermes' tool dispatcher instead of the file and shell tools. With Claude Opus 5.5 driving it, Hermes passed all 9, so the 3B result is about model size, not a verdict on Hermes.
- Hermes installs its Anthropic support from PyPI on first use; offline it fails until pre-installed. Our first Claude run failed every task that way, so we pre-installed it in the image with Hermes' own installer and ran again.
- In the 3B test Ollama capped the context at 32,768 tokens (we asked for 65,536), below the 64K Hermes' docs ask for. Hermes ran anyway without a warning we could see.
- With the local model, every run tried to reach GitHub, OpenRouter and Nous Research hosts (all blocked). With Claude Opus 5.5 it contacted only Anthropic's API.
Measured against Goose
| Measured | Hermes Agent | Goose |
|---|---|---|
| Tasks passed with qwen2.5:3b (of 9) | 0 | 5 |
| Fixed prompt sent with every model call (tokens) | 11,500 | 1,150 |
| Docker image size on disk (GB) | 4.17 | 0.71 |
| Outside hosts it tried to reach | 5 | 1 |
Same model (qwen2.5:3b on 3 CPU cores), same three tasks, three runs each, graded on the files left behind. Ollama capped the context at 32,768 tokens for both, below the 64K Hermes' docs ask for, and a 3B model is far below the model size Hermes asks for, so read the task score as 'which agent copes with a small local model', not overall quality.
Test machine: 8-vCPU AMD EPYC 9354P cloud VM, each test limited to 3 cores and 10 GB RAM, no GPU, no internet (isolated Docker network). Times in seconds are from this CPU-only machine; a desktop with a GPU will be much faster. The VM is shared, so text-generation speed is shown only as a ratio between tools run back to back, not as tokens per second. Docker image size is the on-disk size from docker image ls for every tool. Local AI tools we tested · How we test · raw data
Find your fit
Before you choose Hermes Agent
More options in Agent Frameworks. Compare pricing, platforms, and features side by side.
Brain Memory
Free · Open source
Long-term continuity in multi-session coding projects
Pros
- Memory and saved skills carry over between sessions.
- Reach it from a CLI or messaging apps such as Telegram and Discord.
- You choose where commands run, including Docker, SSH and hosted sandboxes.
- Passed all 9 of our tasks with Claude Opus 5.5 and contacted only Anthropic's API while doing it.
Cons
- Passed 0 of 9 tasks with a 3B local model; it needs a large model and a 64K context.
- Heavy on tokens: an ~11,500-token prompt on every call and about 47,100 tokens per task, six times Goose's.
- It installs its Anthropic support from PyPI on first use, so offline setups fail until you pre-install it.
About
- Platforms
- macOS, Linux, Windows (native or WSL2), Android (Termux); runs commands locally or in Docker, SSH, Singularity, Modal, Daytona, Vercel Sandbox
- API Available
- Yes
- Self-Hosted
- Yes
- Last Updated
- June 9, 2026
Best For
Who it's for
- People who want an assistant that remembers context across sessions
- Teams that turn recurring tasks into reusable skills
What it does well
- Pick up work where you left off, with saved context
- Run scheduled tasks and reach the agent through messaging apps
Integrations
Compare Hermes Agent
Frequently Asked Questions
- Is Hermes Agent free?
- Hermes Agent offers a free option. Hosting, model usage, or optional services may have separate costs; check the pricing details and official terms.
- Is Hermes Agent open source?
- Yes. Hermes Agent is open source.
- Does Hermes Agent have an API?
- Yes. Hermes Agent exposes a developer API. See the official documentation at https://hermes-agent.nousresearch.com for details.
- Can I self-host Hermes Agent?
- Yes. Hermes Agent supports self-hosting on your own infrastructure.
- When was Hermes Agent released?
- Hermes Agent was first released in 2026.
- What platforms does Hermes Agent support?
- Hermes Agent is available on: macOS, Linux, Windows (native or WSL2), Android (Termux); runs commands locally or in Docker, SSH, Singularity, Modal, Daytona, Vercel Sandbox.
People also compared
Curated lists that include this category
Hermes Agent, from Nous Research, is an open-source agent that remembers things between sessions and saves reusable skills. You can reach it from a CLI or through messaging apps like Telegram and Discord.
We ran Hermes Agent v0.22.1 on the same three jobs as Goose and OpenCode: create a file, summarize a report, and count CSV rows with a shell command, three runs each. With a small local model (qwen2.5:3b) it passed none of the 9. That isn't really a verdict on Hermes: the model kept calling Hermes' tool dispatcher instead of the file and shell tools, and Hermes' docs ask for a far bigger model and a 64K context, while our setup gave it 32,768 tokens. With Claude Opus 5.5 it passed all 9. It was also the heaviest of the three agents: a fixed prompt of about 11,500 tokens on every model call, about 47,100 tokens per task, $0.27 for all nine runs, and the slowest time on every task. With the local model it tried to reach GitHub, OpenRouter and Nous Research hosts on every run; with Claude it contacted only Anthropic's API.
Pick Hermes if you want memory, saved skills and messaging access, and you'll pair it with a large model. It's a poor fit for small local models, and it uses far more tokens per task than Goose or OpenCode.
Contribute
- Work on Hermes Agent? Claim this listing.
- Embed a badge for your README or site.
Sign in to rate this tool or suggest an edit.