Connecting AI model providers

Runtype comes with platform keys, so you can start using popular AI models without adding any provider accounts. If you want dedicated rate limits, direct provider billing, or access to a provider’s full model catalog, you can connect your own keys in Settings.

For a full comparison, see Platform Keys vs. Bring Your Own Key (BYOK).

What you get by default

Every Runtype account includes platform keys for popular models from OpenAI, Anthropic, Google, xAI, and open-source models through Mixlayer. You can start building Flows and Agents right away, with no extra setup.

Why connect your own keys?

Adding your own provider keys is optional. It helps when you want dedicated rate limits, direct provider billing, access to the full model catalog, or more headroom for high-volume production workloads.

Supported providers

Runtype supports direct providers such as OpenAI, Anthropic, Google, and xAI; inference providers such as Together.ai, Amazon Bedrock, Vertex AI, and Mixlayer; gateway providers such as Vercel AI Gateway, OpenRouter, Cloudflare Workers AI, and Azure; and any OpenAI-compatible provider via the generic OpenAI-compatible configuration. Models from creators like Mistral, Cohere, AI21, Meta, DeepSeek, and Qwen are accessible through these gateway and inference providers.

Add a provider

You can start either from a model or from a provider.

Option A: Start from a model

1. Go to Settings → Models.
2. Open the My Providers tab.
3. Click Add Model and browse or search the catalog.
4. Select a model. If it needs a key you have not added yet, Runtype prompts you to enter one.

Option B: Start from a provider

1. Go to Settings → Models.
2. Open the My Providers tab.
3. Find the provider card and click Configure.
4. Enter your key details and click Save API Key.
5. Enable the models you want to use.

You can connect multiple providers at once. For example, you might use OpenAI for GPT models and image generation, Anthropic for long-context work, and Google for multimodal use cases, then switch between them across your Flows and Agents.

Get API keys

Each provider has its own setup flow. Use the provider’s dashboard to create the credentials, then add them in Settings → Models under My Providers.

OpenAI

1. Sign in at platform.openai.com.
2. Go to API keys and create a new secret key.
3. Copy the key and paste it into OpenAI → Configure in Runtype.

Anthropic

1. Sign in at console.anthropic.com.
2. Go to API Keys and create a key.
3. Copy the key and paste it into Anthropic → Configure in Runtype.

Google

1. Sign in at aistudio.google.com.
2. Create an API key and copy it.
3. Paste it into Google → Configure in Runtype.

xAI

1. Sign in at console.x.ai.
2. Generate an API key and copy it.
3. Paste it into xAI → Configure in Runtype.

Mixlayer

Mixlayer models are already available through platform keys. If you want dedicated access, create a Mixlayer key and add it in Mixlayer → Configure.

1. Create an account at mixlayer.com.
2. Generate a key from the API keys page.
3. Copy the key and paste it into Runtype.

Amazon Bedrock

1. Create an AWS access key with Bedrock permissions.
2. In Runtype, open Amazon Bedrock → Configure.
3. Enter the access key, secret key, and region.

Vertex AI

1. Create a service account in Google Cloud with Vertex AI permissions.
2. In Runtype, open Vertex AI → Configure.
3. Upload the service account JSON and set the project and location.

OpenAI-compatible providers

Connect any endpoint that speaks the OpenAI API (Groq, Fireworks, Perplexity, Ollama, vLLM, LiteLLM, and more).

1. Get a base URL and API key from your provider.
2. In Runtype, open OpenAI Compatible → Configure.
3. Enter the base URL and API key.
4. Add the models you want available, in one of two ways:
   • Auto-discover: click Discover models and Runtype calls the endpoint’s /v1/models to list what it serves, then you select the ones to enable.
   • Add manually: if the endpoint does not support /v1/models, enter each model id by hand.
5. Click Save to store the key and the model list.

OpenAI-compatible endpoints are a BYOK feature, available on the Growth plan and higher. On other plans the OpenAI Compatible provider appears with a lock and an Upgrade to unlock button. See Platform Keys vs. Bring Your Own Key (BYOK).

Keep your API keys private. Do not share them publicly or commit them to version control.

Choose which models to enable

1. In Settings → Models, click Add Model or open a provider card.
2. Select the models you want to use.
3. Enabled models appear in model pickers across Runtype, including Flows, Agents, and the Playground.

Only enable the models you actively use. This keeps model pickers shorter and easier to scan.

Connect an endpoint programmatically (API, MCP, SDK)

You do not need the dashboard to connect an OpenAI-compatible endpoint. Agents and scripts can do the whole flow — create a key, discover models, and register them — over the API, the MCP server, or the SDK. Every write requires the BYOK entitlement, so on plans without it these calls return 402 with code BYOK_NOT_AVAILABLE. The API key needs the INTEGRATIONS:WRITE scope for provider keys and MODELS:WRITE for model configs.

The flow has three steps: create a generic-openai provider key with the endpoint’s base URL, discover its models (or supply a manual list), then make those models available.

$
# 1. Create the provider key
$
curl -X POST https://api.runtype.com/v1/provider-keys \
>
  -H "Authorization: Bearer YOUR_API_KEY" \
>
  -H "Content-Type: application/json" \
>
  -d '{"provider":"generic-openai","name":"My endpoint","apiKey":"YOUR_PROVIDER_API_KEY","settings":{"baseUrl":"https://api.groq.com/openai/v1"}}'
$
$
# 2. Discover models from the endpoint's /v1/models (returns { "models": [...] })
>
curl -X POST https://api.runtype.com/v1/provider-keys/YOUR_PROVIDER_KEY_ID/discover-models \
>
  -H "Authorization: Bearer YOUR_API_KEY"
>
>
# 3. Make a discrete list of models available
>
curl -X POST https://api.runtype.com/v1/provider-keys/YOUR_PROVIDER_KEY_ID/sync-models \
>
  -H "Authorization: Bearer YOUR_API_KEY" \
>
  -H "Content-Type: application/json" \
>
  -d '{"models":[{"id":"llama-3.1-70b","displayName":"Llama 3.1 70B"}]}'

1
const key = await client.providerKeys.create({
2
  provider: 'generic-openai',
3
  name: 'My endpoint',
4
  apiKey: process.env.PROVIDER_API_KEY,
5
  settings: { baseUrl: 'https://api.groq.com/openai/v1' },
6
})
7
8
// Auto-discover, or skip this and pass your own list to syncModels
9
const { models } = await client.providerKeys.discoverModelsForKey(key.id)
10
await client.providerKeys.syncModels(
11
  key.id,
12
  models.map((m) => ({ id: m.id }))
13
)

Over MCP, the same flow uses the create_provider_key, discover_provider_models, and sync_provider_models tools. To register a single custom model directly, use create_model_config with provider: "generic-openai" and a base_url.

Understand costs with BYOK

When you use your own keys, the provider bills you directly. Runtype does not add extra charges on top. Check your provider’s billing dashboard for exact costs.

Remove a provider

1. Go to Settings → Models and open My Providers.
2. Find the provider and click Manage keys.
3. Remove the key you no longer need and confirm.

If a model is available through platform keys, related Flows and Agents can continue using it after you remove your own key. If the model depends only on your key, update the model selection in those Flows or Agents.

What’s next

• Platform Keys vs. Bring Your Own Key (BYOK) to decide which setup fits your workspace.
• What is the Playground? to test models, prompts, and settings before you build.
• What are Flows? to start using your connected models in repeatable workflows.
• What are Agents? to use your connected models in more autonomous workflows.
• Slack integration to connect Slack as an MCP tool or Surface channel.
• Google Workspace integration to connect Google Workspace services.