Choose and connect AI models

Pulse works out of the box with Oroe AI, the models Oroe runs for you. You never need your own provider. You can still connect one if you want a specific model, like GPT, Claude or Gemini, on your own account.

Everything on this page lives in Settings → AI connections.

What you see on the page

The page reads from top to bottom:

Good to know A model marked Included comes with Pulse. A model from your own provider runs on your own account with that provider.

Connect your own provider

You add a provider in a short dialog with three steps.

1. Open Settings → AI connections.
2. Under Your providers, select Add provider. If you have none yet, the button reads Add your first provider.
3. Pick your provider. The choices are OpenRouter, OpenAI, Anthropic (Claude), Google (Gemini), Groq, Mistral, DeepSeek, Together, Ollama / Beast and Other (OpenAI-compatible).
4. Paste your key into API key. Most providers show a Get a key link if you need one.
5. Check the Name. This is how the provider is listed in Pulse, and you can change it here before you connect.
6. If you are the workspace owner, choose Who can use it: Just me or Everyone in this workspace.
7. Select Connect.

Pulse saves the provider and tests your key straight away. If the test passes, you move on to choosing models.

If the key doesn’t work

Pulse tells you what went wrong in plain words. For example, “the key was refused”, “the account is out of credit” or “the address looks wrong”. It also shows what the provider said.

The provider is saved but not working yet. Select Try a different key to paste a new one, or close the dialog and fix it later from the provider’s row.

Self-hosted and other endpoints

Ollama / Beast and Other (OpenAI-compatible) ask for an Address, such as https://…/v1. For the listed providers, the address is filled in for you. You only change it through Use a different address.

Note Addresses must use https. Plain http, private network addresses and local addresses are blocked unless an administrator has allow-listed them. If yours is blocked, Pulse says “this address is blocked — an administrator must allow it”.

Choose models

A provider can offer hundreds of models. You pick the few you want Pulse to use.

1. After connecting, the dialog shows Choose models from your provider. You can also get there later with Choose models, or with Browse all models from the provider’s ⋯ menu.
2. Select Select recommended for a good starting set, or search and filter the list yourself. Filters include Tools, Vision, Reasoning, Free and Long context ≥128k.
3. Select the register button to add the models you chose. It shows how many you picked, for example Register 3 models. Registering calls nothing and costs nothing.
4. On the last step, select Use for chat if you want chat to use one of your new models. Then select Done.

If a provider does not publish a list of its models, you add one by its exact id instead. Open the provider’s row, expand Advanced, and use Add a model by its id.

Each model shows whether it passed a check, for example Works, including tools or Not checked yet. Select Check to test one model, or Details to see what it can do.

Set the model chat uses

1. Under What Pulse uses, select Change on the Chat line.
2. If you are the workspace owner, choose who This default applies to: Everyone in this workspace or Just me. Everyone else sets a personal default.
3. Pick a model from the Model list. Models are grouped under Oroe AI and Your connections.

The choice saves straight away. You can also switch models for a single conversation from the chat header. See Chatting with Pulse.

Document search

The Document search line shows the model Pulse uses to search your documents. Pulse picks it automatically, so your documents are searchable without any setup. There is no workspace-wide switch for it.

Routing and spending limits

For more control over chat, open Advanced settings and select Edit routing and limits…. From there you can:

• Choose Fixed model, where every step runs on the one model you choose, or Optimized routing, where Pulse chooses among the models you list and never outside them.
• Set a Budget for one task (optional), with a Token limit, Request limit or Spending limit.
• Decide what Pulse does with models that have no configured price.

Select Save policy when you’re done.

Look after a provider

Each provider row has a status, such as Ready, Not checked or Needs attention. The ⋯ menu offers:

• Test connection checks that the key still works.
• Browse all models adds more models.
• Rotate key replaces the key. The old one stops working here immediately.
• Remove provider removes the provider and deletes its saved key.

Note When you remove a provider, anything using its models, including chat, shows a message rather than quietly switching to another provider. Afterwards, choose another chat model under What Pulse uses.

To change a provider’s name or address, remove it and add it again.

How your keys are kept

Keys are encrypted on the server and are write-only. Pulse can store, rotate and remove a key, but it can never show it back to you or to anyone else.

Tip If you used an older version of Pulse that kept your key in the browser, you may see Move your provider key into Pulse. Select Import so the key works on all your devices. Pulse removes the browser copy only after the server confirms it has it.

Who can do what

• Anyone can add a provider for their own use.
• Only the workspace owner can share a provider with everyone, or set the chat default for the whole workspace.
• If a provider was shared with you, you can use and test it. Managing it stays with its owner.

Next: Chatting with Pulse.