How to Migrate from OpenClaw to Hermes
If you're moving from OpenClaw to Hermes, you're in luck: Hermes ships a dedicated migration command for exactly this
If you're moving from OpenClaw to Hermes, you're in luck: Hermes ships a dedicated migration command for exactly this. It's one of the clearest signals in the agent landscape — when a tool builds a named path off a specific competitor, it tells you where it sees itself winning.
Here's how the migration works, what carries over, and the one thing no migration tool moves for you.
How do you migrate from OpenClaw to Hermes?
Run hermes claw migrate. It's a first-class command in Hermes, listed alongside core setup commands — Nous Research built a deliberate path off OpenClaw.
The migration imports:
- Conversation history — brought into Hermes's searchable session store
- Skills — OpenClaw's
SKILL.mdfiles converted to Hermes's skill format - Memory files —
MEMORY.mdand daily notes imported into Hermes's memory system - Workspace configuration — mapped to Hermes's config format
Worth knowing: migration paths now exist in both directions. Hermes ships hermes claw migrate, and OpenClaw has documented importing Hermes state through a bundled migration path — so which is "easier" depends on version and direction. Either way, treat this as tool-state migration: files, skills, and history can move, but the agent's inferred understanding of you doesn't fully transfer with them.
What transfers and what doesn't?
The files transfer cleanly. Your accumulated notes, your skills, your history — all of it lands in Hermes in a usable form. For an OpenClaw user with an established setup, that's the hard part handled.
What doesn't transfer is everything that wasn't written to a file. OpenClaw's memory is only ever as complete as what the agent logged — and the agent decides what's worth saving, inconsistently. So the migration faithfully moves what OpenClaw managed to capture, along with all the gaps in what it didn't. You inherit the same partial picture, now in a new tool.
And critically, Hermes's signature feature — the model of you it builds over time — doesn't come across, because it never existed in OpenClaw to migrate. Which leads to the part most migration guides skip.
Why does Hermes still feel like it doesn't know you after migrating?
Because Hermes learns who you are by watching you work, and migration doesn't give it that history. The whole premise of Hermes is a compounding model of you built session by session. A freshly migrated Hermes has your old files but none of that accumulated understanding — it's a capable agent that, in the way that matters most, still meets you cold.
So you get the odd experience of a successful migration that still feels like starting over. The skills are there; the knowing you isn't. You're back to the weeks-long process of the agent gradually inferring your role, your preferences, your priorities from behavior.
How do you bring your context with you?
Seed Hermes with structured context instead of waiting for it to learn you.
Rather than letting the agent slowly reconstruct who you are, give it an accurate, structured profile from the first session — your role, background, current work, how you operate. Hermes starts informed and its learning loop builds on a correct foundation instead of guessing its way to one.
The deeper point: migrating between agents is the exact problem portable context solves once and for all. If your context lives in a layer outside any single tool, switching from OpenClaw to Hermes — or running both, or adding Claude next — doesn't mean rebuilding it each time. The agent changes; your context stays put and loads into whatever you're using. Migration stops being a context-loss event and becomes just a tool swap.
You migrated the files. Let a context layer handle the part the files can't carry.
→ How Hermes memory works: How Does Hermes Memory Work?
→ The portable-context approach: How to Move Your Context from ChatGPT to Claude
