Custom endpoints
Any OpenAI-compatible API — local LLMs, gateways, Azure OpenAI, etc.
No coding required. If you have never used an API key before, follow the walkthrough below step by step. Developers can skip to Quick reference at the bottom
Part 1 — Get your API key
- This path is for technical users running Ollama, LiteLLM, vLLM, Azure OpenAI, etc.
- Find the base URL your server exposes (often ending in /v1)
- Find the model name string your server expects
- Copy an API key if your gateway requires one (Ollama locally often uses any placeholder)
Part 2 — Add your connection in Airbox
- Dashboard → Connections → + LLM provider
- Enter type: openai_compat
- Paste API key if required
- When asked for model, enter your model string (e.g. llama3)
- You will also be prompted for base URL — e.g. http://localhost:11434/v1 for local Ollama
Open the dashboard in a new tab if you want to follow along
Part 3 — Route your inbox
Your Airbox address does not use a connection until you attach one:
- In the Inboxes card, find your Airbox address
- Click the gear icon (⚙) on that address to open routing settings
- Under "AI, LLM or MCP provider", choose the connection you just added from the dropdown
- Leave "Email a reply back" checked if you want results in your inbox (recommended)
- Click "Save"
Try it — send an email
Subject: (* #summarize *)
New to commands? See copy-paste examples and the syntax cheat sheet
Quick reference
| Dashboard type | openai_compat |
| Email targets | @custom |
| Default model | (you choose) |
| Auth | From your server admin or local tool docs |
Short checklist
- openai_compat + base_url + model
What you get
- Self-hosted and corporate gateways
- Full URL and model control
