Skip to content

LM Studio

LM Studio is a desktop application for running local and remote LLMs, with millions of downloads. Version 0.3.5 introduced MCP support via Streamable HTTP, allowing local models to call Picsart tools during a conversation.

Prerequisites

  • LM Studio 0.3.5 or later (download)
  • A local model loaded that supports tool/function calling (see model recommendations below)
  • A Picsart API key, available from picsart.com/ai-playground/ under API settings

Setup

  1. Open LM Studio.
  2. Click the Settings icon (gear, top right).
  3. Select the MCP tab.
  4. Click Add MCP Server.
  5. Fill in the following fields:
    • Name: Picsart Gen AI
    • Type: Streamable HTTP
    • URL: https://mcp.picsart.io/mcp
  6. Under Headers, add:
Authorization: Bearer YOUR_PICSART_API_KEY
  1. Click Save. LM Studio connects to the server and lists available Picsart tools.
  2. Load a model with strong tool-calling support in the model selector.
  3. Open the chat and send: List available Picsart models.

For the complete MCP reference, see the LM Studio MCP documentation.

Tool-calling reliability varies by model. The following models produce consistent results with Picsart MCP:

  • Qwen 2.5-7B-Instruct or larger
  • Mistral-7B-Instruct-v0.3 / Mistral Nemo
  • Llama 3.1 8B Instruct or larger

Models not fine-tuned for tool use may ignore tool calls or format them incorrectly.

Use it

With a tool-capable model loaded, send prompts that reference Picsart capabilities:

Generate a product image with Flux 2 Pro and return the URL.
Use Picsart to remove the background from https://example.com/photo.jpg.
How many Picsart credits do I have left?

The model decides when to invoke a Picsart tool based on your prompt. Results are returned inline in the chat.

Troubleshooting

MCP tab not visible in Settings

Upgrade LM Studio to version 0.3.5 or later. Older versions do not include MCP support.

Tool calls fail intermittently

Switch to a model with stronger function-calling support. Qwen 2.5-7B-Instruct and Mistral-7B-Instruct-v0.3 are reliable starting points. If the issue persists with a supported model, reload the model from the selector.

"Connection refused" or server timeout

LM Studio may be blocked from making outbound requests by a firewall or VPN. Confirm that your machine can reach mcp.picsart.io on port 443. If you are on a corporate network, check proxy settings.

Authorization error (401)

Verify that the header is entered exactly as Authorization: Bearer YOUR_PICSART_API_KEY with a single space after Bearer and no extra characters. Regenerate the key from picsart.com/ai-playground/ if needed.

FAQ

Can LM Studio use Picsart MCP with any local model?

Technically yes, but results vary. Models not fine-tuned for tool use may ignore tool calls or produce malformed requests. Qwen 2.5, Mistral, and Llama 3.1 are the most reliable choices.

Does this cost Picsart credits even though I am running a local model?

Yes. The local model handles conversation and decides when to call a tool, but the actual generation (image, video, audio) runs on Picsart's servers and consumes credits from your account.

Can I use Picsart MCP alongside other MCP servers in LM Studio?

Yes. LM Studio supports multiple MCP servers simultaneously. Add each server separately in the MCP tab.

Does LM Studio support streaming responses from Picsart tools?

Picsart tool responses return a result URL once generation is complete. There is no streaming mid-generation output. LM Studio displays the result when the tool call finishes.

Start creating

The Picsart MCP server is now connected. Visit the documentation for examples, available models, and prompt ideas.

Ready to generate?

View documentation

Built on @picsart/ai-sdk · gen-ai CLI · Picsart MCP · Skills