At the beginning, I treated Hermes like a smart chat assistant: ask a question, get an answer, maybe have it write some code. But the more you use it, the more obvious the limitation becomes.

I ran Hermes for a week and came back to a disk full of garbage. Logs, test scripts, browser profiles, random artifacts. I had no idea what was safe to delete. Spent an hour digging through folders. Then I found the Disk Cleanup plugin.

That little headache taught me something: Hermes Agent is not just another chat assistant. Its real power comes from the way you can extend it. Plugins add tools, hooks, integrations, dashboards, model providers, memory systems, and image generation backends without touching the core agent. Bundled plugins ship with Hermes and live in plugins/. Memory providers use a standard interface for advanced storage. The plugin system is opt-in by default, so you enable only what you actually need.

Why does that matter? Most users don't want an agent that only talks. They want one that cleans up files, joins meetings, traces sessions, generates images, remembers context, connects to apps, and fits into their actual workflow.

Here are the best Hermes plugins and plugin-style extensions worth knowing in 2026.

Quick Reference

Rank	Plugin	Type	One-Liner
1	Disk Cleanup	Built-in	Stops temp files piling up silently
2	Langfuse Observability	Built-in	Turns Hermes from a black box into something you can inspect
3	Composio Connect	Third-party MCP	Opens the door to 1,000+ SaaS apps via OAuth
4	Kanban Dashboard	Built-in	Structure for multi-agent work that survives crashes
5	Google Meet	Built-in	Brings Hermes into live meetings
6	Honcho Memory Provider	Memory provider	Builds a model of how you think, not just what you said
7	Hindsight Memory Provider	Memory provider	Knowledge-graph memory with structured entity retrieval
8	Hermes Achievements	Dashboard plugin	Usage insights surfaced from your actual session history
9	Hermes Agent Self-Evolution	External Framework	Runs optimization loops for skills and prompts
10	Hermes Web Search Plus	Third-party Search	Better search when research is a repeat workflow

Before You Install Anything

Hermes has a few different extension types, so it helps to separate them.

Some plugins ship with Hermes and can be enabled directly:

hermes plugins enable <plugin-name>

Some capabilities live as optional skills:

hermes skills install official/<category>/<skill-name>

Memory providers use the memory setup flow:

hermes memory setup

Third-party MCP integrations usually have their own install path, often through a URL, config entry, or external service.

That distinction matters because not everything on this list is a simple one-command plugin. Some are built in. Some are memory providers. Some are external frameworks that work alongside Hermes rather than inside it.

To uninstall any skill:

hermes skills uninstall <skill-name>

Full reference: hermes-agent.nousresearch.com/docs/reference/cli-commands

Now let’s get into the actual list.

1. Disk Cleanup

Best for: Anyone who lets Hermes touch files

Type: Built-in plugin

Docs: hermes-agent.nousresearch.com/docs/user-guide/features/overview

How to install

It ships with Hermes. Enable it with:

hermes plugins enable disk-cleanup

What it is

Agents create a lot of temporary files. Test scripts, scratch outputs, browser profiles, logs, generated artifacts, one-off experiments. Left alone, those pile up fast and you end up doing what I did: spending an hour deciding what's safe to delete.

Disk Cleanup tracks temp files created during agent sessions and gives you slash commands to clean them safely.

/disk-cleanup status     # see what's accumulated
/disk-cleanup dry-run    # preview what would be removed
/disk-cleanup quick      # remove obvious temp files
/disk-cleanup deep       # thorough cleanup pass

The dry-run is the important one. You see the full list before anything gets touched.

Pros

Solves a boring but real problem that gets worse the longer you run Hermes. The dry-run mode means you're never flying blind. Works silently in the background once configured.

Cons

Won't catch everything. Files created outside of Hermes's tracked scope (things you created manually in the same directory) aren't flagged. It's not a general disk management tool, just a cleanup layer for agent-generated artifacts.

Why install it

Any agent that works with files needs this. Without it, the workspace quietly fills up and you eventually spend time you don't have sorting through folders trying to figure out what Hermes created versus what you need.

2. Langfuse Observability

Best for: Developers, teams, debugging agent workflows

Type: Built-in observability plugin

Docs: hermes-agent.nousresearch.com/docs/user-guide/features/overview

Langfuse: langfuse.com

How to install

pip install langfuse
hermes plugins enable observability/langfuse

Then add your keys to ~/.hermes/.env:

LANGFUSE_PUBLIC_KEY=your-public-key
LANGFUSE_SECRET_KEY=your-secret-key
LANGFUSE_HOST=https://cloud.langfuse.com  # or your self-hosted URL

What it is

If you use Hermes for anything serious, observability becomes important quickly. The Langfuse plugin sends Hermes activity into Langfuse's tracing dashboard: turns, LLM calls, tool invocations, latency, token usage, cost per session.

That makes it much easier to answer questions that are otherwise genuinely hard: which tool calls actually happened, where did the agent spend time, which model calls were expensive, why did a task fail halfway through, what was the exact sequence of steps during a long autonomous run.

I would install this early if you are building workflows or spending real money on model calls. For casual use, it is useful but not urgent.

Pros

Turns a black box into something you can inspect. For developers testing complex agent workflows, the trace view is invaluable. Cost tracking alone justifies the install if you're running Hermes with any frequency. Langfuse is also self-hostable if you don't want traces going to a cloud service.

Cons

Adds a small latency overhead per turn as traces are sent to Langfuse. If you're on the free tier there are trace retention limits. Not useful for casual single-user setups where you're just chatting, but for anything production-facing or team-facing it's the first plugin you should enable.

Why install it

For developers, teams, or anyone testing agent workflows, this is the difference between guessing what Hermes did and actually knowing.

3. Composio Connect

Best for: Connecting Hermes to everyday SaaS tools

Type: Third-party MCP integration

Docs: composio.dev/hermes

Toolkits: composio.dev/toolkits

How to install

Paste this URL directly into a Hermes chat session:

<https://composio.dev/hermes>

Hermes installs it and walks you through connecting your apps via OAuth. No config file editing, no credential copying.

What it is

Out of the box, Hermes's integrations are limited to what the core system natively supports. Composio Connect opens the door to 1,000+ applications via MCP: Slack, Gmail, Notion, Linear, Jira, HubSpot, GitHub, Google Calendar, Salesforce, and hundreds more. All authentication is OAuth, which means you're not pasting API keys anywhere.

The tool router handles everything through a single MCP endpoint, so your Hermes config stays clean instead of accumulating individual server entries for each app.

Some things it enables:

Turn a Slack message into a Linear ticket automatically
Pull calendar context before Hermes drafts an email
Update HubSpot from agent output without touching the CRM
Chain actions across apps that don't have native Hermes integrations

Pros

By far the fastest way to expand what Hermes can actually reach. OAuth means no credential management. 1,000+ apps through one integration point is a different scale from anything else in this list. The install path is one of the simplest of any plugin here.

Cons

Your OAuth tokens flow through Composio's infrastructure. For personal productivity most people won't care. For teams handling sensitive business data it's worth reading the security documentation first: composio.dev/security

Why install it

Without this, Hermes can't reach most of the tools your team actually uses. With it, the gap between "agent that responds to prompts" and "agent that does things in your actual work environment" closes significantly.

4. Kanban Dashboard

Best for: Multi-agent workflows and long-running projects

Type: Built-in dashboard plugin

Docs: hermes-agent.nousresearch.com/docs/user-guide/features/web-dashboard

How to install

pip install 'hermes-agent[web,pty]'
hermes dashboard

This starts a local server at http://127.0.0.1:9119. The Kanban tab is available inside the dashboard. For the full install with all extras:

pip install 'hermes-agent[all]'

Then load the supporting skills:

hermes skills load devops/kanban-orchestrator
hermes skills load devops/kanban-worker

What it is

The Kanban Dashboard gives Hermes a SQLite-backed task board shared across all profiles on a host. Tasks carry an assignee, optional dependency links, a workspace kind, and a tenant namespace. The dashboard includes drag-and-drop boards, run history, worker logs, live WebSocket updates, and per-task event attribution.

For multi-agent work or long-running projects, a visual board is much easier to follow than scrolling through a chat transcript to figure out what happened.

Pros

SQLite backing means tasks survive crashes, restarts, and reboots. The live WebSocket updates mean you're not refreshing to see progress. Runs entirely on localhost, nothing leaves your machine. The orchestrator and worker skills pair directly with the dashboard for a complete task delegation setup.

Cons

For single-task casual use, this is overkill. The value shows up when you're coordinating multiple agent workers or tracking work across sessions. The standalone daemon mode is deprecated; make sure you're running it through the gateway.

Why install it

When multi-agent workflows get complex enough that a chat transcript is no longer a useful way to track progress, this is what you reach for.

5. Google Meet

Best for: Meeting notes, transcripts, and action items

Type: Built-in standalone plugin

Docs: hermes-agent.nousresearch.com/docs/user-guide/features/overview

How to install

hermes plugins enable google-meet

Configure your speech-to-text provider in ~/.hermes/config.yaml:

voice:
  stt_provider: whisper   # or groq, openai, faster-whisper

What it is

The Google Meet plugin lets Hermes join meetings as a headless participant, transcribe live audio through your chosen speech-to-text provider, and generate artifacts from the session: transcripts, structured notes, action items.

This is one of the most immediately practical plugins for teams. Hermes shows up to the standup, takes notes, and has the action items ready in Notion or Linear before the call is over.

Pros

Works with six STT providers: local faster-whisper (free, runs on-device), Groq, OpenAI Whisper, Mistral, xAI, and a local CLI wrapper. The local faster-whisper option means transcription stays on your machine. Chains naturally with the Notion and Linear bundled skills to push meeting output directly into your project tools.

Cons

Meeting transcription needs consent. Worth being explicit with your team that Hermes is in the call. Transcription quality on fast speakers or heavy accents depends on which STT provider you're running. The plugin is headless, so it doesn't handle Zoom or Teams natively; Google Meet is the main supported platform.

Why install it

It gives Hermes a role in real-time team workflows rather than just async ones. Standups, research calls, interviews, project syncs, client calls. The notes show up automatically without anyone having to write them.

6. Honcho Memory Provider

Best for: A more personal Hermes over time

Type: Memory provider plugin (optional skill)

Docs: hermes-agent.nousresearch.com/docs/user-guide/features/honcho

GitHub: github.com/plastic-labs/honcho

API keys: honcho.dev

How to install

hermes memory setup
# select "honcho" from the provider list

Or manually in config:

# ~/.hermes/config.yaml
memory:
  provider: honcho

echo 'HONCHO_API_KEY=your-key' >> ~/.hermes/.env

Also install the optional skill for full configuration access:

hermes skills install official/autonomous-ai-agents/honcho

What it is

Hermes's built-in memory is file-based: MEMORY.md and USER.md that the agent maintains across sessions. That works fine. Honcho is what you install when fine isn't enough.

Instead of storing raw text, Honcho builds a model of the user through dialectic reasoning. After each conversation turn, it analyzes the exchange and derives insights about preferences, habits, communication style, and goals. These accumulate over time. The agent's understanding of you deepens the more you use it, not just because it stores more, but because it reasons about what it stores.

It also handles multi-agent isolation properly: if you run a coding assistant and a personal assistant as separate Hermes profiles, Honcho keeps their user models separate so they don't contaminate each other's context.

Pros

Dialectic reasoning produces a qualitatively different kind of memory than raw storage. The cold/warm prompt selection (explained in the docs) means it calibrates differently at the start of a session vs. mid-conversation. Three independent tuning knobs for context refresh rate, dialectic frequency, and reasoning depth. Full semantic search across the memory history.

Cons

Requires an API key from honcho.dev. The dialectic reasoning adds latency per turn, tunable via dialecticCadence. Takes a few sessions to build up useful context; day-one value is low. More to configure than the default memory system.

Why install it

Default Hermes memory remembers facts. Honcho builds a model of how you think. If you want an agent that gets meaningfully more personalized over months of use rather than just accumulating stored text, this is the memory provider for that.

7. Hindsight Memory Provider

Best for: Precise, structured memory retrieval

Type: Memory provider plugin

Docs: hermes-agent.nousresearch.com/docs/user-guide/features/memory-providers

How to install

hermes memory setup
# select "hindsight" from the provider list
# leave the API key blank for local daemon

Or manually in config:

# ~/.hermes/config.yaml
memory:
  provider: hindsight

What it is

Most memory providers store semantic chunks and retrieve by similarity. Hindsight does something structurally different: it extracts discrete facts, named entities, and the relationships between them, then stores those as a knowledge graph.

Ask about a specific service and Hindsight surfaces facts about that service. Not chunks that happened to mention it nearby. Actual structured knowledge about it.

The reflect operation is what makes it unusual. It's a periodic synthesis pass that reads across all stored memories, derives higher-level insights, and consolidates related facts. Other providers accumulate raw extractions. Hindsight actively refines what it knows over time.

Pros

Highest retrieval precision of any memory provider for specific entity queries. The knowledge graph structure means relationships between things are preserved, not just individual facts. Local daemon mode means nothing leaves your machine if privacy matters. The reflect operation means stored knowledge improves over time rather than just growing.

Cons

More complex to configure than Honcho or the default memory system. Knowledge graph extraction adds overhead compared to raw semantic storage. Better suited for people with real privacy requirements or precision retrieval needs than for casual use.

Why install it

Developers who need the highest retrieval accuracy, or anyone handling data they'd prefer not to send to a cloud memory service. For everything else, Honcho is probably the better starting point.

8. Hermes Achievements

Best for: Understanding your own usage patterns

Type: Dashboard plugin

Docs: hermes-agent.nousresearch.com/docs/user-guide/features/extending-the-dashboard

How to install

pip install 'hermes-agent[web]'
hermes dashboard

Achievements appear as a tab inside the web dashboard automatically once the dashboard is running.

What it is

Hermes Achievements adds a tab to the dashboard with badges earned from your actual Hermes session history. The plugin scans local sessions and unlocks achievements based on real behavior: autonomous tool chains, debugging and recovery patterns, file edits, skill usage, memory usage, cron jobs, plugin usage, web research, and various usage patterns.

Most achievements level through five tiers: Copper, Silver, Gold, Diamond, and Olympian. The catalog has over 60 achievements.

It's not the most critical plugin on this list, but it surfaces something genuinely useful: visibility into how you actually use the agent versus how you think you use it.

Pros

Turns usage history into something readable. Good for noticing patterns, habits, and underused parts of Hermes you've set up but forgotten about. The achievement framing makes it less dry than a raw analytics dashboard.

Cons

Purely a visibility tool. Adds nothing to agent capability. More useful after a few months of sessions than on day one when there's nothing to surface yet.

Why install it

Running hermes dashboard for the first time and seeing the Achievements tab is a good moment. It's a different angle on how much work the agent has actually done.

9. Hermes Agent Self-Evolution

Best for: Advanced users improving skills and prompts

Type: External optimization framework

Docs: github.com/NousResearch/hermes-agent-self-evolution

How to install

Install the self-evolution engine as a separate package that works alongside your main Hermes Agent. This is not a normal built-in plugin.

git clone <https://github.com/NousResearch/hermes-agent-self-evolution.git>
cd hermes-agent-self-evolution
pip install -e ".[dev]"
export HERMES_AGENT_REPO=~/.hermes/hermes-agent

Then run an optimization loop against a specific skill:

python -m evolution.skills.evolve_skill --skill github-code-review --iterations 10 --eval-source synthetic

What it is

Self-Evolution is an external optimization framework for improving Hermes skills, prompts, and tool descriptions. Instead of manually editing a skill every time it performs badly, you can run an evaluation loop and let the framework generate candidate improvements.

That sounds powerful, and it is, but it is not magic. You still need to inspect the result, understand what changed, and decide whether the new version is actually better.

This is not something I would recommend as a first-week Hermes install. It belongs later, after you have a skill that fails in repeatable ways and you can tell whether the new version is actually improving.

Pros

Useful for maintaining skills you rely on often. Gives you a more systematic way to improve prompts and tool descriptions instead of guessing. Good fit for developers who already have evaluation criteria or test cases for agent behavior.

Cons

Not plug-and-play. Bad evaluations can produce bad improvements. A prompt can get better at the test while becoming worse in real use. Costs also depend on the model, number of iterations, and evaluation setup, so each run should be treated like a small experiment.

Why install it

Install this if you are already building or maintaining Hermes skills and want a more structured way to improve them. Skip it if you are still learning Hermes or mostly using the agent for one-off tasks.

10. Hermes Web Search Plus

Best for: Better research workflows

Type: Third-party search enhancement

Docs: github.com/robbyczgw-cla/hermes-web-search-plus

How to install

Clone the plugin into your Hermes plugin directory and install the required Python dependencies.

git clone <https://github.com/robbyczgw-cla/hermes-web-search-plus.git> ~/.hermes/plugins/web-search-plus

cd ~/.hermes/hermes-agent
source venv/bin/activate

pip install requests
cd ~/.hermes/plugins/web-search-plus
cp .env.template .env

Then enable the plugin tools in ~/.hermes/config.yaml:

plugins:
  enabled:
    - web-search-plus

tools:
  enabled:
    - web
    - web-search-plus

Restart Hermes, or use /restart in chat. The new tools web_search_plus and web_extract_plus should then be available.

What it is

Hermes has a built-in web_search tool, but research-heavy workflows often need more than a basic search command. Web Search Plus adds provider routing, extraction support, fallback behavior, and caching, depending on how you configure it.

It is best understood as a research upgrade, not just another search command. The value is not that it magically finds perfect sources. The value is that it gives Hermes more ways to search, extract, and recover when one provider is not the right fit.

I would not install this just because it exists. If you only need occasional lookups, the built-in search may be enough. This becomes useful when research is a repeated workflow and bad search results are actually slowing you down.

Pros

Can improve research quality when configured well. Provider routing is the main reason to use it, especially if your queries vary between quick lookups, technical docs, and broader market research. Caching can also help avoid repeated API calls during similar searches.

Cons

Setup is more involved than the built-in search tool. You may need API keys for the providers you want to use. More providers also means more configuration, more rate limits, and more places where something can fail. Better search also does not replace source judgment, so Hermes still needs to evaluate what it finds.

Why install it

Install this if Hermes is part of your research workflow and basic search keeps getting in the way. Skip it if you only need occasional web lookups.

Which Plugin Is Best for What?

If you want...	Start here
A clean, manageable workspace	Disk Cleanup
To actually see what Hermes is doing	Langfuse Observability
Hermes to reach your SaaS apps	Composio Connect
Multi-agent task tracking that survives reboots	Kanban Dashboard
Meeting notes without writing them	Google Meet
Memory that builds a model of you over time	Honcho
Precise structured knowledge retrieval	Hindsight
Visibility into your usage history	Hermes Achievements
A way to improve Hermes skills systematically	Hermes Agent Self-Evolution
Better research workflows than basic search	Hermes Web Search Plus

The Best Stack for Most Users

You do not need all ten.

For most people, the strongest starting stack is:

Disk Cleanup + Langfuse + Composio Connect

That gives you a cleaner workspace, traceability, and access to the apps where your work actually happens.

If you use Hermes daily, add memory:

Honcho for personalization
Hindsight for precision

If you run complex projects, add:

Kanban Dashboard

If meetings are part of your workflow, add:

Google Meet

If research is a serious use case and built-in search is slowing you down, add:

Web Search Plus

If you are already building or maintaining Hermes skills, Self-Evolution is worth exploring. Do not make it one of your first installs.

Bottom Line

The best Hermes plugins are not the ones that sound the most futuristic.

They are the ones that remove the annoying parts of using an agent every day.

Disk Cleanup keeps the workspace sane. Langfuse shows you what actually happened. Composio connects Hermes to the tools where real work lives. Kanban makes long-running work easier to follow. Google Meet turns conversations into usable output. Honcho and Hindsight make memory more serious.

That is what makes Hermes interesting in 2026. Not just that it can answer questions, but that you can shape it into something closer to an actual operating layer for your work.

Start with the boring plugins first. They are usually the ones you will still be using three months later.