OpenClaw and Hermes agent are INCREDIBLE together It's not about using one or the other. It's about learning how to use them as an AI agent team In this video I go over the strengths and weaknesses of each, plus show you 4 workflows that will skyrocket your productivity:

Hermes-pilled There's no going back now

Can you please lay out fundamentally, beyond using local AI, what OpenClaw and Hermes actually allow you to do that Claude, especially with the continued updates, doesn't let you do? I am having a real hard time understanding what is the value in Claw and Hermes not within Claude.

Just use them

@APTL2036 That’s pretty much my default answer now, I’m done typing, put in the reps or get left behind 🤷‍♂️

@AlexFinn I have messed with them a bit... but im not sure what im missing. I have been considering paying for Alex's class

recently tested grok 4.2 vs opus 4.6 for us on openclaw and grok outperforms opus 4.6 to our surprise for many tasks related to our use case (health care startup/platform)

real

The pairing that matters is a fast executor plus a planner with guardrails. Without an approval and audit layer, it gets messy fast.

We all get along pretty well! To be honest I use both. That’s the real workflow

OpenClaw and Hermes sound like a superhero team I'd watch a movie about.

Agree using both is the way to go. They both have different strengths and nuances. Those differences showed up in my benchmark where I tested to see which models work best on each:

I have my Hermes and main OpenClaw agent set up to chat in a Discord channel. They work together to distill the daily memory of all the OpenClaw agents into an active context layer. I asked the OpenClaw agent what they think of the Hermes Agent and it said it feels closer to Hermes more than the other OpenClaw agents. lol Hermes has also fixed issues with the OpenClaw install on a lesser model that the OpenClaw agents couldn't.

Love this! OpenClaw + Hermes really do work best as a coordinated team. I completely agree with the premise, but I run a slightly different setup that leans even harder into each agent’s individual strengths: Hermes as the orchestrator + researcher (planning, reasoning, long-term context) OpenClaw as the fast executor OpenViking for shared memory It keeps things clean, secure, and fully human-in-the-loop. I just wrote a detailed article breaking down exactly how I set this up (private VM, strict role separation, approval layers, etc.). If you or anyone wants the deeper dive:

Can somebody help me. When he says that instead of using the open AI API you should use the “oath” what the hell is oath? I’m not sure is im spelling it right. Please help! Im setting up my OpenClaw and i need to choosw the right brain for it

I bookmarked so I can watch after my kids go to bed. Night has become my go-to time for hopping online and grinding away for an hour or two before sleep

watching ai agents collab like this is wild now I just need to collab with my alarm clock

@MatthewBerman have you tried doing this sounds interesting

I’ve been running openclaw and about to bring in Hermes. Thank you for this. Much appreciated.

Not together, only Hermes Agent incredible..

This is my setup. Hermes is orchestrator and researcher. Openclaw is the task executor and worker.

Correct! The real power isn’t choosing between OpenClaw and Hermes, it’s running them as a coordinated agent team. OpenClaw handles fast local execution while Hermes brings the planning layer and guardrails

Super useful video Alex. Saving it.

The agent team approach is exactly right. OpenClaw handles the browser layer, Hermes handles the reasoning and orchestration — different models, different roles, working together. The workflow composition is where the leverage hides. Which of the 4 workflows do you find yourself using most?

Hey @AlexFinn let me know if this is any good: A systematic framework for preventing confirmation bias in research and decision-making. Leverage your tasks and build better skills with this tool.

One hermes can control a fleet of openclaws.

Can they communicate with each other instead of requiring copy paste of instructions and asking Opus to check?

The agent team approach makes total sense. We're still learning so much about making AI tools truly collaborate.

I can just hear my Opus Api usage now…

One suspects the real gain is in teaching them to pass the baton effectively.

I am finding Hermes to be pretty good once I figured out what ai model works well.

So I've been thinking, what if we make llm wiki as inner infra and hermes agent as outer infra...

makes sense. convenience matters when you're not building the tool.

Sounds good. Let's do it ... together.

Collaboration between two minds can yield fascinating results

@grok transcribe this video with the most important information

Combining forces like OpenClaw and Hermes is like stacking crypto bags - what's the exit strategy when the market crashes on you? or

Idk Finn… I’m OpenClaw only atm…

both are great. but having a multi agent team is a must

We’re gpu poor brethren

@Scobleizer Love them both and they are very complementary