Switchboard keys start with swb_. Create and revoke them in Switchboard account settings. The plaintext key is shown once on create, so copy it into your environment immediately.

Create your key

Add the package to your Package.swift:

dependencies: [
    .package(url: "https://github.com/valni-labs/switchboard-swift", from: "0.6.0")
],
targets: [
    .target(name: "MyApp", dependencies: [
        .product(name: "Switchboard", package: "switchboard-swift"),
    ])
]

The library product is Switchboard. The SDK lives at valni-labs/switchboard-swift, and valni-labs/SwitchboardDemo is a clone-and-run macOS app showing a full integration.

Set SWITCHBOARD_API_KEY in your environment, then:

import Switchboard

guard let apiKey = ProcessInfo.processInfo.environment["SWITCHBOARD_API_KEY"] else {
    fatalError("Set SWITCHBOARD_API_KEY in the environment")
}
let client = Client(apiKey: apiKey)

let response = try await client.inference(Inference.Request(
    model: "claude-sonnet-5",
    messages: [
        .system("You are a senior Swift engineer."),
        .user("Write a one-liner that flattens [[Int]] into [Int]."),
    ],
    user: "end-user-id"
))

print(response.content ?? "")

One endpoint serves every model in the catalog:

curl https://switchboard.valni.app/v1/switchboard/inference \
  -H "Authorization: Bearer $SWITCHBOARD_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-5",
    "messages": [{"role": "user", "content": "Hello, Switchboard."}],
    "user": "end-user-id"
  }'

Pass any catalog model ID. Switchboard handles upstream routing, retries on 429, and failover across provider keys.

• claude-sonnet-5: frontier general-purpose
• claude-opus-4-8: frontier agent / long-form
• gpt-5.5: frontier general-purpose
• deepseek-v4-flash: baseline, cheapest
• kimi-k2.6: long-context

The full live catalog is one call away: client.models() in the SDK, or GET /v1/switchboard/models with your key.

• See pricing. Alpha is invite-only while we size the provider key pool.
• Email hello@valni.ai with feedback, missing models, or alpha-access requests.