Using LiteLLM with OpenClaw

Prerequisites

• ✓An existing OpenClaw setup where you can run the `openclaw` CLI.
• ✓Python and pip available so you can install and run the LiteLLM proxy.
• ✓Network access from your OpenClaw environment to `http://localhost:4000` or wherever LiteLLM runs.
• ✓API keys for the underlying providers you plan to route through LiteLLM (for example `ANTHROPIC_API_KEY` and `OPENAI_API_KEY`).

Steps

1. 1

   Run OpenClaw onboarding with LiteLLM auth

   If you are starting from a fresh OpenClaw project, use the onboarding flow so OpenClaw knows it should authenticate via LiteLLM. This wires up the provider choice without you editing config by hand.

   bashCopy

   openclaw onboard --auth-choice litellm-api-key

2. 2

   Install and start the LiteLLM proxy

   Run the LiteLLM proxy so OpenClaw has a single HTTP endpoint to send all model traffic to. This example starts LiteLLM with `claude-opus-4-6` exposed; you can expand this later with more models and routing.

   bashCopy

   pip install 'litellm[proxy]'
   litellm --model claude-opus-4-6

3. 3

   Export your LiteLLM API key for OpenClaw

   OpenClaw reads the LiteLLM API key from the `LITELLM_API_KEY` environment variable. Set this in the same shell or process environment where you run `openclaw` so every request is authenticated against LiteLLM.

   bashCopy

   export LITELLM_API_KEY="sk-litellm-key"

4. 4

   Start OpenClaw with LiteLLM as the model provider

   Once LiteLLM is running and the API key is exported, start OpenClaw normally. With the LiteLLM auth choice and env var in place, OpenClaw routes its model calls through the LiteLLM proxy.

   bashCopy

   export LITELLM_API_KEY="your-litellm-key"
   
   openclaw

5. 5

   Configure LiteLLM as a provider in your OpenClaw config file

   Define LiteLLM as a model provider in your OpenClaw config so you can reference specific LiteLLM models like `litellm/claude-opus-4-6` and `litellm/gpt-4o`. This also sets the default agent model to the LiteLLM-backed Claude model.

   jsonCopy

   {
     models: {
       providers: {
         litellm: {
           baseUrl: "http://localhost:4000",
           apiKey: "${LITELLM_API_KEY}",
           api: "openai-completions",
           models: [
             {
               id: "claude-opus-4-6",
               name: "Claude Opus 4.6",
               reasoning: true,
               input: ["text", "image"],
               contextWindow: 200000,
               maxTokens: 64000,
             },
             {
               id: "gpt-4o",
               name: "GPT-4o",
               reasoning: false,
               input: ["text", "image"],
               contextWindow: 128000,
               maxTokens: 8192,
             },
           ],
         },
       },
     },
     agents: {
       defaults: {
         model: { primary: "litellm/claude-opus-4-6" },
       },
     },
   }

6. 6

   Create a LiteLLM virtual key dedicated to OpenClaw

   Use LiteLLM virtual keys to give OpenClaw its own API key with a monthly budget cap. This lets you control and monitor OpenClaw’s spend separately from other clients hitting the same LiteLLM proxy.

   bashCopy

   curl -X POST "http://localhost:4000/key/generate" \
     -H "Authorization: Bearer $LITELLM_MASTER_KEY" \
     -H "Content-Type: application/json" \
     -d '{
       "key_alias": "openclaw",
       "max_budget": 50.00,
       "budget_duration": "monthly"
     }'

7. 7

   Set up LiteLLM model routing for multiple backends

   yaml` so model names like `claude-opus-4-6` and `gpt-4o` route to the right upstream providers. OpenClaw keeps requesting the same model IDs while LiteLLM swaps or fans out to Anthropic and OpenAI under the hood.

   yamlCopy

   model_list:
     - model_name: claude-opus-4-6
       litellm_params:
         model: claude-opus-4-6
         api_key: os.environ/ANTHROPIC_API_KEY
   
     - model_name: gpt-4o
       litellm_params:
         model: gpt-4o
         api_key: os.environ/OPENAI_API_KEY

8. 8

   Inspect LiteLLM usage and spend for your OpenClaw traffic

   Once OpenClaw is sending traffic through LiteLLM, use the LiteLLM dashboard APIs to inspect key info and spend logs. This helps you confirm that the right key is in use and that your budget limits behave as expected.

   bashCopy

   # Key info
   curl "http://localhost:4000/key/info" \
     -H "Authorization: Bearer sk-litellm-key"
   
   # Spend logs
   curl "http://localhost:4000/spend/logs" \
     -H "Authorization: Bearer $LITELLM_MASTER_KEY"

Configuration

Troubleshooting

pip install 'litellm[proxy]'
litellm --model claude-opus-4-6

export LITELLM_API_KEY="sk-litellm-key"

Frequently asked questions