Guide
Use Your Ollama Cloud API Key with Hosted Hermes Agent
Save your Ollama key once on the API keys page, pick an Ollama Cloud model for your Hermes bot, and chat on your own Ollama plan. No config files, no custom endpoint.
Checked on 9 October 2026. Ollama's model list and plan limits change, so the live list in the dashboard is the one to trust.
What the Ollama Cloud card does
The API keys page has an Ollama Cloud API Key card. Once you save a key there, your Hermes bots can use Ollama Cloud models directly. Requests go from your bot to https://ollama.com/v1 with your key, so usage is billed by Ollama on your plan. This works on Hermes Agent bots only for now.
Before this card existed, the only route was the generic Custom API Provider card. That still works, but the dedicated card is simpler. It checks the key when you save it, lists Ollama's current models, and keeps every Hermes bot on Ollama Cloud updated when you change or remove the key.
Set it up in three steps
- Create an API key at ollama.com/settings/keys. A free Ollama account is enough to start.
- Open openclawlaunch.com/api-keys, find the Ollama Cloud API Key card, paste the key and click Save. The card saves the key, shows Active with a masked preview, and sends Ollama a one-token test request. If that test fails, a warning appears under the card.
- On the dashboard, open the model picker on your Hermes bot, choose Ollama Cloud in the left column, and pick a model. The bot restarts in the background for a few seconds and then answers through Ollama.
The card itself also has a model picker with a Set as primary model button. That button switches every running Hermes bot on your account at once. To change one bot only, use that bot's own picker on the dashboard.
Which models you can pick
The list comes live from Ollama, so new models appear without an update on our side. On 9 October 2026 it had 18 text models, including gpt-oss:120b, gpt-oss:20b, gemma4:31b, glm-5.3, kimi-k2.6, deepseek-v4.1-flash and minimax-m2.7.
| Ollama plan | Models that answered | Limit to know |
|---|---|---|
| Free, no credits bought | gpt-oss:120b, gpt-oss:20b, gemma4:31b, nemotron-3-nano:30b | Starter credits, 1 request at a time |
| Bought credits, or Pro and above | All listed models, including glm-5.3 and kimi-k2.6 | Usage credits; Pro and above also allow more parallel requests |
gpt-oss:120b is preselected because it works on the free plan. Plan details are on Ollama's pricing page.
Changing or removing the key
Saving a new key updates every running Hermes bot that is on an Ollama Cloud model. Click Remove to delete the key. Any Hermes bot that was on an Ollama Cloud model then moves back to the default model, which runs on your plan credits again, or on your own OpenRouter key if you saved one. If no default is available, the bot stops answering until you pick another model for it.
Common problems
- "Ollama did not accept it" after Save: the key is wrong or was revoked, or Ollama could not be reached at that moment. Click Remove, check the key, and paste it again. If the warning stays, create a fresh key at ollama.com/settings/keys. Removing the key moves Ollama Cloud bots off their model, so pick it again afterwards.
- A model replies with a payment or credits error: that model needs Ollama usage credits. Switch to gpt-oss:120b, or buy credits in your Ollama account.
- Slow or failed replies during busy tasks: the free plan allows one request at a time, so parallel subagents or overlapping tasks can be slowed or turned away. A paid plan allows more at once.
- Ollama Cloud is missing from the picker: the bot is an OpenClaw bot, or the key has not been saved yet. The section appears on Hermes bots once the card shows Active.
Frequently asked questions
Does this work on OpenClaw bots?
Not yet. The Ollama Cloud card works with Hermes Agent bots only. OpenClaw ships its own Ollama provider that handles Ollama addresses itself, so the hosted key path is not wired for it. To run Ollama Cloud with a self-hosted OpenClaw, see the OpenClaw + Ollama Cloud guide.
Do Ollama Cloud chats use my OpenClaw Launch credits?
No. Once a bot is on an Ollama Cloud model, its chat requests go straight to Ollama with your key, so they count against your Ollama plan instead.
Can I use the Ollama free plan?
Yes, for Ollama's starter models. On 9 October 2026 a free key could chat with gpt-oss:120b, gpt-oss:20b, gemma4:31b and nemotron-3-nano:30b. Larger models such as glm-5.3, kimi-k2.6 and deepseek-v4.1-flash need usage credits, which you can buy on any Ollama plan or get monthly with Pro and above.
Is this the same as self-hosting Hermes with Ollama Cloud?
The model calls are the same, but here you never edit config.yaml or run hermes model. The dashboard writes the provider settings for you. For the self-hosted flow, read Hermes Agent + Ollama Cloud.