Docs Standalone Kubernetes Tutorials Standalone Kubernetes Blog Enterprise Community Get Started GitHub

Anthropic

Configure Anthropic (Claude) as an LLM provider in agentgateway.

Before you begin

Install and set up an agentgateway proxy.

Set up access to Anthropic

  1. Get an API key to access the Anthropic API.

  2. Save the API key in an environment variable.

    export ANTHROPIC_API_KEY=<insert your API key>

  3. Create a Kubernetes secret to store your Anthropic API key.

    kubectl apply -f- <<EOF
apiVersion: v1
kind: Secret
metadata:
  name: anthropic-secret
  namespace: agentgateway-system
type: Opaque
stringData:
  Authorization: $ANTHROPIC_API_KEY
EOF

  4. Create an AgentgatewayBackend resource to configure your LLM provider that references the Anthropic API key secret.

    kubectl apply -f- <<EOF
apiVersion: agentgateway.dev/v1alpha1
kind: AgentgatewayBackend
metadata:
  name: anthropic
  namespace: agentgateway-system
spec:
  ai:
    provider:
      anthropic:
        model: "claude-haiku-4-5-20251001"
  policies:
    auth:
      secretRef:
        name: anthropic-secret
EOF

    Review the following table to understand this configuration. For more information, see the API reference.

    Setting Description
    ai.provider.anthropic Define the LLM provider that you want to use. The example uses Anthropic.
    anthropic.model The model to use to generate responses. In this example, you use the claude-haiku-4-5-20251001 model.
    policies.auth Provide the credentials to use to access the Anthropic API. The example refers to the secret that you previously created. The token is automatically sent in the x-api-key header.

  5. Create an HTTPRoute resource that routes incoming traffic to the AgentgatewayBackend. The following example sets up a route on the /anthropic path. Note that agentgateway automatically rewrites the endpoint to the Anthropic /v1/messages endpoint.

    kubectl apply -f- <<EOF
apiVersion: gateway.networking.k8s.io/v1
kind: HTTPRoute
metadata:
  name: anthropic
  namespace: agentgateway-system
spec:
  parentRefs:
    - name: agentgateway-proxy
      namespace: agentgateway-system
  rules:
  - backendRefs:
    - name: anthropic
      namespace: agentgateway-system
      group: agentgateway.dev
      kind: AgentgatewayBackend
EOF
    kubectl apply -f- <<EOF
apiVersion: gateway.networking.k8s.io/v1
kind: HTTPRoute
metadata:
  name: anthropic
  namespace: agentgateway-system
spec:
  parentRefs:
    - name: agentgateway-proxy
      namespace: agentgateway-system
  rules:
  - matches:
    - path:
        type: PathPrefix
        value: /v1/chat/completions
    backendRefs:
    - name: anthropic
      namespace: agentgateway-system
      group: agentgateway.dev
      kind: AgentgatewayBackend
EOF
    kubectl apply -f- <<EOF
apiVersion: gateway.networking.k8s.io/v1
kind: HTTPRoute
metadata:
  name: anthropic
  namespace: agentgateway-system
spec:
  parentRefs:
    - name: agentgateway-proxy
      namespace: agentgateway-system
  rules:
  - matches:
    - path:
        type: PathPrefix
        value: /anthropic
    backendRefs:
    - name: anthropic
      namespace: agentgateway-system
      group: agentgateway.dev
      kind: AgentgatewayBackend
EOF

  6. Send a request to the LLM provider API along the route that you previously created. Verify that the request succeeds and that you get back a response from the API.

    Cloud Provider LoadBalancer:

    curl "$INGRESS_GW_ADDRESS/v1/messages" -H content-type:application/json  -d '{
   "model": "",
   "messages": [
     {
       "role": "user",
       "content": "Explain how AI works in simple terms."
     }
   ]
 }' | jq

    Localhost:

    curl "localhost:8080/v1/messages" -H content-type:application/json  -d '{
   "model": "",
   "messages": [
     {
       "role": "user",
       "content": "Explain how AI works in simple terms."
     }
   ]
 }' | jq

    Cloud Provider LoadBalancer:

    curl "$INGRESS_GW_ADDRESS/v1/chat/completions" -H content-type:application/json  -d '{
   "model": "",
   "messages": [
     {
       "role": "user",
       "content": "Explain how AI works in simple terms."
     }
   ]
 }' | jq

    Localhost:

    curl "localhost:8080/v1/chat/completions" -H content-type:application/json  -d '{
   "model": "",
   "messages": [
     {
       "role": "user",
       "content": "Explain how AI works in simple terms."
     }
   ]
 }' | jq

    Cloud Provider LoadBalancer:

    curl "$INGRESS_GW_ADDRESS/anthropic" -H content-type:application/json  -d '{
   "model": "",
   "messages": [
     {
       "role": "user",
       "content": "Explain how AI works in simple terms."
     }
   ]
 }' | jq

    Localhost:

    curl "localhost:8080/anthropic" -H content-type:application/json  -d '{
   "model": "",
   "messages": [
     {
       "role": "user",
       "content": "Explain how AI works in simple terms."
     }
   ]
 }' | jq

    Example output:

    {
  "model": "claude-haiku-4-5-20251001",
  "usage": {
    "prompt_tokens": 16,
    "completion_tokens": 318,
    "total_tokens": 334
  },
  "choices": [
    {
      "message": {
        "content": "Artificial Intelligence (AI) is a field of computer science that focuses on creating machines that can perform tasks that typically require human intelligence, such as visual perception, speech recognition, decision-making, and language translation. Here's a simple explanation of how AI works:\n\n1. Data input: AI systems require data to learn and make decisions. This data can be in the form of images, text, numbers, or any other format.\n\n2. Training: The AI system is trained using this data. During training, the system learns to recognize patterns, relationships, and make predictions based on the input data.\n\n3. Algorithms: AI uses various algorithms, which are sets of instructions or rules, to process and analyze the data. These algorithms can be simple or complex, depending on the task at hand.\n\n4. Machine Learning: A subset of AI, machine learning, enables the system to automatically learn and improve from experience without being explicitly programmed. As the AI system is exposed to more data, it can refine its algorithms and become more accurate over time.\n\n5. Output: Once the AI system has processed the data, it generates an output. This output can be a prediction, a decision, or an action, depending on the purpose of the AI system.\n\nAI can be categorized into narrow (weak) AI and general (strong) AI. Narrow AI is designed to perform a specific task, such as playing chess or recognizing speech, while general AI aims to have human-like intelligence that can perform any intellectual task.",
        "role": "assistant"
      },
      "index": 0,
      "finish_reason": "stop"
    }
  ],
  "id": "msg_01PbaJfDHnjEBG4BueJNR2ff",
  "created": 1764627002,
  "object": "chat.completion"
}

Connect to Claude CLI

Configure your AgentgatewayBackend resource to allow connections to the Claude Code CLI.

Keep the following things in mind:

  • Model selection: If you specify a specific model in the AgentgatewayBackend resource and then use a different model in the Claude Code CLI, you get a 400 HTTP response with an error message similar to thinking mode isn't enabled. To use any model, remove the spec.ai.provider.anthropic.model field and replace it with {}.
  • Routes: To use the Claude Code CLI, you must explicitly set the routes that you want to allow. By default, the Claude Code CLI sends requests to the /v1/messages API endpoint. However, it might send requests to other endpoints, such as /v1/models. To ensure that the Claude Code CLI forwards these requests to Anthropic accordingly without using the /v1/messages API, add a * passthrough route to your AgentgatewayBackend resource as shown in this guide.

  1. Update your AgentgatewayBackend resource to allow connections to the Claude Code CLI. The following example sets the default /v1/messages and a catch-all passthrough API endpoints, and allows you to use any model via the Claude Code CLI.

    kubectl apply -f- <<EOF
apiVersion: agentgateway.dev/v1alpha1
kind: AgentgatewayBackend
metadata:
  name: anthropic
  namespace: agentgateway-system
spec:
  ai:
    provider:
      anthropic: {}
  policies:
    ai: 
      routes:
        '/v1/messages': Messages
        '*': Passthrough
    auth:
      secretRef:
        name: anthropic-secret
EOF

  2. Test the connection via the Claude Code CLI by sending a prompt.

    Run the Claude Code CLI with a prompt:

    ANTHROPIC_BASE_URL="http://$INGRESS_GW_ADDRESS:80" claude -p "What is a credit card"

    Start the Claude Code CLI terminal and start prompting it:

    ANTHROPIC_BASE_URL="http://$INGRESS_GW_ADDRESS:80" claude

    Run the Claude Code CLI with a prompt:

    ANTHROPIC_BASE_URL="http://localhost:8080" claude -p "What is a credit card"

    Start the Claude Code CLI terminal and start prompting it:

    ANTHROPIC_BASE_URL="http://localhost:8080" claude

    Example output:

    A credit card is a payment card issued by a financial institution (typically a bank) that allows the cardholder to borrow funds to pay for goods and services, with the agreement to repay the borrowed amount, usually with interest.

## Key characteristics:

**How it works:**
- The issuer extends a **credit limit** — the maximum you can spend
- You make purchases on credit (borrowed money)
- You receive a monthly statement
- You can pay the full balance or a minimum payment

**Costs:**
- **APR (Annual Percentage Rate):** Interest charged on unpaid balances, typically 15-30%
- **Annual fee:** Some cards charge a yearly fee
- **Late fees:** Charged if you miss payment deadlines

**Benefits:**
- Build credit history/score
- Purchase protections and fraud liability limits
- Rewards (cashback, points, miles)
- Emergency purchasing power

**Key difference from a debit card:**
- Debit cards draw directly from your bank account (your money)
- Credit cards use borrowed money you repay later

**Risks:**
- Debt accumulation if balances aren't paid in full
- High interest charges
- Potential negative impact on credit score if mismanaged

In short: a credit card is a short-term loan instrument that, when used responsibly, offers convenience and benefits, but can become costly if balances carry over month-to-month.

Next steps

Amazon BedrockAzure OpenAI
Agentgateway assistant

Ask me anything about agentgateway configuration, features, or usage.

Note: AI-generated content might contain errors; please verify and test all returned information.

↑↓ navigate ↵ select esc dismiss

What could be improved?