Connect an integration
Summary
This step completes when you create an API key for the customer-facing REST API: the credential that lets another tool in your stack pull CyberHQ.ai data programmatically.
Before you start
You need the Manage Tenant permission: without it, Settings → Integrations → API keys doesn't appear. This key is distinct from any personal user-profile key you may already have: it authenticates the platform-to-platform integration itself, not you personally. Available on every plan.
1. Create the key
- Go to Integrations → API keys.
- Give it a name that identifies what it's for (the tool or workflow it connects to, not just "API key 1").
- Choose its scopes: the specific data and actions it's allowed to access.
- Optionally set a rate limit, and an expiry date: the key stops working at the end of that day.
- Create the key.
2. Copy it immediately
The full key value is shown exactly once, at the moment of creation. Copy it into whatever tool is going to use it right away: after this screen, you'll only ever see a short prefix, not the full value again.
3. Use it in the connecting tool
Paste the key into the receiving tool's configuration wherever it asks for API credentials. Exactly where that is depends on the tool: this step is about confirming the key works, not about any one integration's setup.
Confirm it worked
Make one request with the key from the connecting tool (or a simple test call, if the tool provides one) and confirm it returns data rather than an authentication error.
Set your AI inference policy and provider
Go to Configure → Integrations → AI provider. Two things are set here.
AI inference policy governs every AI task in the platform, and has three settings:
| Setting | What it means |
| Disabled | No AI. AI-backed features fall back to built-in rules. |
| CyberHQ.ai built-in model | AI tasks run on models provided with CyberHQ.ai (AWS Bedrock), with no API key of your own. |
| External Allowed | AI tasks use your own provider API key (for example Anthropic). |
Model configuration is where you choose the provider: Anthropic (Claude) using your own API key, AWS Bedrock, or Ollama (self-hosted). When no provider is connected the page reads "No AI provider in use".
Check the data residency line before connecting. The page states where inference runs, for example Australia, under a Data residency heading; with your own Anthropic key it reads Data residency (if you connect AWS Bedrock) instead. If your organization has data-residency obligations, read this before choosing a provider.
To stop using a provider, disconnect it and confirm in the Disconnect provider dialog. The key (or Bedrock configuration) and every per-task model choice are removed, and AI features such as Iris and the briefing narrative stop working until you connect a provider again and pick the models.
If it didn't work
- Everything email-delivered has stopped: check Integrations → Email provider. The state "Per-tenant provider configured (email delivery is off)" looks configured but sends nothing.
- An AI model won't save: access is verified when you save, not while you browse. The page says "Showing default models; access is verified when you save.", so a model appearing in the list isn't proof your provider account can use it.
- "The AI provider is unreachable right now, try again shortly.": a transient fault on the provider connection. Retry; if it persists, raise it with support.
- "Self-hosted model unreachable": the endpoint is configured but not responding. "No self-hosted endpoint is configured for this deployment." means the option isn't available on your deployment at all; choose another provider or contact support.
- Authentication fails immediately: the key may not have been copied in full, or a scope required for the request you're making wasn't selected at creation.
- Lost the key before pasting it anywhere: there's no way to retrieve the full value again. Rotate it to get a new value, or revoke it and create a new one. Either way the old value stops working immediately, so anything depending on it breaks until you update it.
- Need broader or narrower access later: rather than editing scopes on an existing key, create a new one with the scopes you need and revoke the old one once the new one is confirmed working, so nothing depending on it breaks mid-transition.
Related
- Complete your organization profile
- Get to know Support