> ## Documentation Index
> Fetch the complete documentation index at: https://docs.open-hive.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Credentials
> Connect the accounts, tools, and models your agents work with.
**Configuration → Credentials** is where you give agents access to the outside world.
It has two halves: **Built-in Connections** for accounts OpenHive integrates with
directly, and **API Keys** for the long tail of everything else.
Search across both from the top of the page.
## Built-in Connections
These are first-class integrations, authorized over OAuth rather than by pasting a key.
| Connection | Unlocks |
| ----------- | ------------------------------------------------------------------------------------------------------------- |
| **Google** | Gmail, Calendar, Drive, Docs, and Sheets. One Google grant covers every OpenHive tool that hits a Google API. |
| **GitHub** | Repos, issues, pull requests, and actions. Fine-grained and classic personal access tokens both work. |
| **HubSpot** | Contacts, deals, marketing emails, and pipelines. Used by sales queens that update HubSpot directly. |
| **Notion** | Connects through an internal integration token. |
A connection can hold **multiple accounts** — use **Add another account** when you need
separate identities for separate workspaces or clients.
### Alert channels
**Telegram** and **Slack** are connected here too, but they carry alerts rather than
data. Connect one and you get pinged when a colony stalls — and you can reply in the
chat to keep it moving, without opening the app.
Alert channels are enabled per colony, in that colony's **Automations** tab, not
globally. Connecting the channel here only makes it available.
## API Keys
Everything else is an API key. The catalog is large — several hundred entries covering
CRMs, enrichment and data providers, warehouses, databases, and email delivery. Each
entry lists exactly the fields it needs, from a single key to a full connection (host,
port, database, user, password).
Sort by **Most used** to surface the common ones, or search if you know what you want.
Several entries link to **How to get key**, which saves hunting through a provider's
docs.
The catalog is broader than what OpenHive can currently complete a connection for.
Entries that aren't wired up yet show **Connect** disabled, so you can see what's
coming without starting a flow that can't finish.
## Model providers
Your model keys live in the same API Keys section. By default agents run on OpenHive's
routed model pool, drawing on your bundled allotment; adding your own provider key
bypasses that allotment and bills you directly with the provider.
| Provider | Notes |
| -------------- | ------------------------------------------------------------------------------------- |
| **Anthropic** | Claude Opus, Sonnet, and Haiku. |
| **DeepSeek** | V3 and R1 reasoning models. Cheap throughput for bulk tool-using turns. |
| **Groq** | LPU-backed inference for sub-second responses — Llama and Mixtral at high throughput. |
| **Moonshot** | Kimi K2 with 1M+ token context. Good for long-document and multi-file agents. |
| **OpenRouter** | Open-source model proxy — Llama, Qwen, DeepSeek, Mixtral and more behind one key. |
| **OpenHive** | The routed pool. Used as the fallback when no other provider key is configured. |
For high-volume work — bulk enrichment, large prospect lists — a DeepSeek or Groq key
is usually the cheapest way to scale it. See
[Credits and plans](/account/credits).
Credentials are held for use by the runtime on your machine. Treat the machine running
OpenHive as you would any machine holding production API keys.