Enable Your Models to Collaborate

Install MCP servers with one click and let your LM Studio models query each other, work in parallel, and access external tools.

๐Ÿ”„

Concurrent Multi-Agent

Production

Query multiple models simultaneously with true async operations and connection pooling. Perfect for comparing responses or getting diverse perspectives.

  • Query up to 10 models concurrently
  • Connection pooling for efficiency
  • Automatic retry on failures
  • 1.18x faster than sequential
๐Ÿ”

Code Reviewer

Production

Automated code review using 30B+ models. Analyzes code for security issues, performance problems, and best practices.

  • Security vulnerability detection
  • Performance optimization tips
  • Architecture analysis
  • Batch file processing
๐Ÿ—„๏ธ

Oracle Database

Production

Connect your LM Studio models directly to Oracle databases. Execute queries, analyze schemas, and generate reports.

  • Execute SQL queries
  • Schema exploration
  • Table analysis
  • Connection pooling
๐Ÿƒ

MongoDB

Production

Perform MongoDB operations directly from your models. Query collections, aggregate data, and manage documents.

  • CRUD operations
  • Aggregation pipelines
  • Collection management
  • Index optimization
๐ŸŽญ

Agent Orchestrator

Beta

Coordinate multiple specialized agents with different roles. Automatically route tasks to the best model for each job.

  • Role-based agent assignment
  • Task routing logic
  • Agent collaboration
  • Result aggregation
๐Ÿ’พ

VRAM Manager

Alpha

Automatically manage model loading and unloading to stay within VRAM limits. Run more models than your GPU can hold.

  • JIT model loading
  • Auto-unload unused models
  • VRAM monitoring
  • 33GB limit enforcement

๐Ÿ“Š Model Requirements

Choose the right model size for your use case

Task Complexity Minimum Model VRAM Usage Best For
Simple Tasks qwen2.5-coder-14b 8.5 GB Basic function calls, single tool operations, simple automation
Medium Complexity codestral-22b 13.5 GB Multi-step workflows, sequential logic, conditional operations
High Complexity qwen3-coder-30b 18 GB Nested workflows, dynamic tool selection, complex reasoning
Expert Level qwen3-32b 19.5 GB Error handling, retry logic, production-grade systems