Skip to main content

Integrations

The Integrations tab connects b4 to outside services. AI Assistant lets b4 ask an AI provider about its settings, while MCP server lets an AI application ask b4. Changes are saved with Save Changes in the page header.

IPinfo​

FieldWhat it doesDefault
Access tokenThe token for ipinfo.io. With it, the Add to a set dialog of the Traffic page and the dashboard offers Enrich with IPInfo, which shows the organization, host name and location of an address. b4 requests these details from the router. Get a token opens ipinfo.io, where the free tier is enough.Empty

AI Assistant​

The assistant explains settings inside the web interface. b4 sends the question to the chosen AI provider. While the assistant is on, an AI icon shows next to many fields, most of them in the set editor, and Ask AI to explain this setting opens the answer.

FieldWhat it doesDefault
Enable AI assistanceThe switch in the card header.Off
OpenAI, Anthropic, Ollama, OpenAI-compatibleThe provider. With OpenAI or Anthropic, the questions and the parts of the configuration they touch leave the router for that provider. With Ollama they stay local while Endpoint points to the router or a computer in the local network.None
ModelThe model, picked from the provider's list or typed in.Empty
API keyRequired for OpenAI and Anthropic, optional for OpenAI-compatible. Set key stores it on the router in a file of its own, not in the configuration.Empty

The Advanced section holds the connection details.

FieldWhat it doesDefault
EndpointThe address of the provider's API, filled in for each provider. Another value points to another server, such as a self-hosted one.Per provider
Max tokensThe longest answer, in tokens.1024
TemperatureFrom 0 to 2. Lower values give more predictable answers.0.2
Timeout (s)How long one request may take.120

While the provider cannot be used, for example without a key, the card shows the reason.

MCP server​

The MCP server lets an AI application, such as Claude Desktop, Cursor or LM Studio, read the state of b4 and, when allowed, change its settings. The model runs in that application, so b4 needs no API key for it. Connecting an AI assistant shows the setup.

FieldWhat it doesDefault
Enable MCP serverThe switch in the card header. The server uses the port of the web interface.Off
Access tokenThe token the AI application presents. Generate creates one, Regenerate replaces it. Without a token, the application needs a web login while one is set, and that login ends when b4 restarts. Without either, anyone who reaches the port can use the server, including changes and probes when they are allowed.Empty
Allow configuration changesLets the AI change the sets and the MTProto, SOCKS5 and logging settings, not only read them. Credentials and all other settings stay read-only.Off
Allow active probesLets the AI make b4 fetch a domain and run Discovery, which sends real traffic from the router.Off
Client configurationThe entry to paste into the AI application, with the address and the token.-

Community Hub​

The connection to the Community Hub, where b4 users publish sets and report whether they work. Community Hub describes the card in full.

FieldWhat it doesDefault
Enable Community HubAdds the Community page, the hourly sync and the Share with other b4 users section of the set editor.Off
Hub mirrorsAddresses of the same hub, one per line, tried in order. Empty means https://hub.b4core.app.Empty
Hub public keyOnly for a self-hosted hub with its own key. The key of hub.b4core.app is built in.Empty

Hub status shows the catalogue, the last sync with its error and the hub addresses, and Sync now syncs at once. Author identity shows the key that signs the sets and reports sent from this router. Show recovery code gives the code that moves this key to another router, where Restore from code takes it in.