AI connections & profiles
Every AI feature in Helpin runs through two building blocks: connections, which hold the credentials for an AI provider, and profiles, which choose the model. Together they decide which provider and model an agent uses, and how that usage is paid for.
Manage both in Settings → AI.
Connections
A connection stores encrypted credentials for one provider.
Provider | Notes |
|---|---|
OpenAI | API key |
Anthropic | API key |
OpenRouter | API key. One key gives access to many model vendors |
OpenAI-compatible endpoints | Self-hosted or private models your administrator has approved |
ChatGPT | Sign in with a ChatGPT account, where enabled. Always personal |
Connections come in two kinds:
Shared connections belong to the workspace. They can run unattended work, such as the support agent, automations, and scheduled agents. Managing them requires workspace settings permission.
Personal connections belong to you. Use them when you start work manually, for example in Ask Agent.
Profiles
A profile binds a connection to a specific model and its settings. Every workspace starts with four standard profiles:
Profile | Use it for |
|---|---|
Small | Fast, inexpensive everyday work. Ask Agent and new custom agents start here |
Medium | A balance of quality and cost |
Large | Harder reasoning and planning |
Flagship | The most capable model, for the hardest problems |
You can edit the standard profiles or create your own. A profile can also name one fallback profile, which is used only if its primary connection is unavailable before a run starts.
Which profile a run uses
Helpin picks the first of these that's set:
A profile you choose manually when starting the run
The agent's default profile
The workspace's default profile
Once a run starts, it keeps its provider and model through retries and resumes. Helpin never switches models in the middle of a run. Changing a profile later doesn't affect runs that have already started.
Paying for AI usage
How AI is paid for | |
|---|---|
Helpin Cloud | Usage comes out of your plan's monthly AI allowance, at the rates on the pricing page. Optional overage is billed monthly |
Helpin Cloud, Enterprise | You can also bring your own provider keys |
Self-hosted (Community) | Always your own keys. Helpin never charges for AI usage |
Self-hosting? See AI providers & knowledge search for server keys and knowledge embeddings.
Disconnecting a provider
Disconnecting a connection deletes its stored secret and stops new model calls on runs that use it. Requests already in flight may finish. Reconnect with a new key to resume work that paused for authentication.
Was this article helpful?