LM Studio API Configuration Guide

Overview

LM Studio is a local LLM model manager that runs models on your machine and exposes an OpenAI-compatible API.

Key features:

• Local-first inference (data stays on your machine)
• OpenAI-compatible interface (easy integration with existing tools)
• No API key required by default
• Built-in GUI for loading and managing local models

Model identifier mapping in PicoClaw:

LM Studio API Identifier	PicoClaw Model Identifier	Notes	Use Cases
openai/gpt-oss-20b	lmstudio/openai/gpt-oss-20b	From LM Studio right-side panel	Daily conversations and coding
your-model-id	lmstudio/your-model-id	Works with any loaded model	Custom local workflows

---

Getting API Key

Step 1: Open LM Studio Local Server

In LM Studio, go to Developer -> Local Server.

Step 2: Start Server and Load Model

1. Ensure Status is Running
2. Click Load Model and load the model you want to use
3. Confirm the local endpoint (default: http://localhost:1234/v1)

Step 3: Copy API Model Identifier

1. Copy API Model Identifier from the right panel
2. Use it in PicoClaw as lmstudio/<identifier>

Note: LM Studio does not require API Key by default. Only set API Key in PicoClaw if your LM Studio server has authentication enabled.

---

Configuring PicoClaw

Option 1: Using WebUI (Recommended)

PicoClaw provides a WebUI interface where you can easily configure models without manually editing configuration files.

Edit preset settings, or click the "Add Model" button in the top right corner:

Field	Value
Model Alias	Custom name, e.g., lmstudio-local
Model Identifier	lmstudio/openai/gpt-oss-20b (replace suffix with your identifier)
API Key	Leave empty by default (only needed when auth is enabled)
API Base URL	Leave empty (default: http://localhost:1234/v1)

Option 2: Edit Configuration File

Add in config.json:

{
  "model_list": [
    {
      "model_name": "lmstudio-local",
      "model": "lmstudio/openai/gpt-oss-20b"
    },
    {
      "model_name": "lmstudio-lan",
      "model": "lmstudio/deepseek-r1-distill-llama-8b",
      "api_base": "http://10.20.30.40:1234/v1"
    }
  ],
  "agents": {
    "defaults": {
      "model": "lmstudio-local"
    }
  }
}

PicoClaw strips the lmstudio/ prefix before sending requests. If api_base is not set, it defaults to http://localhost:1234/v1.

---

Limits & Quotas

Billing

LM Studio local inference does not have per-token cloud billing. Costs mainly come from your local hardware usage.

Rate Limits

• Throughput depends on your CPU/GPU and model size
• Larger models or higher parallel settings increase latency
• LAN/remote access also depends on network stability

---

Common Issues

Cannot Connect to Local Server

Cause: LM Studio local server is not running, wrong port, or firewall restrictions

Solutions:

• Verify Status is Running
• Check endpoint and port in LM Studio
• Ensure local firewall/network policy allows access

Model Not Found

Cause: Model is not loaded, or model identifier mismatch

Solutions:

• Load the model in LM Studio first
• Re-copy API Model Identifier and use lmstudio/<identifier>

401 Unauthorized

Cause: LM Studio authentication is enabled, but no token is provided

Solution:

• Set API Key in PicoClaw model configuration