# API Documentation
Complete reference for the Local LLM MCP Server tools, resources, and prompts.
## Table of Contents
- [Tools](#tools)
- [Resources](#resources)
- [Prompts](#prompts)
- [Model Management](#model-management)
- [Error Handling](#error-handling)
---
## Tools
### `local_reasoning`
Use local LLM for general-purpose reasoning and question answering.
**Parameters:**
- `prompt` (string, required): The question or task to process
- `system_prompt` (string, optional): System context or role definition
- `model_params` (object, optional): Model parameters
- `temperature` (number, 0-2): Randomness (default: 0.7)
- `max_tokens` (number, 1-8192): Maximum response length
- `top_p` (number, 0-1): Nucleus sampling
- `frequency_penalty` (number, -2 to 2): Penalize repeated tokens
- `presence_penalty` (number, -2 to 2): Penalize new topics
- `model` (string, optional): Specific model name (overrides default)
**Example:**
```json
{
"prompt": "Explain quantum entanglement in simple terms",
"system_prompt": "You are a physics teacher explaining concepts to beginners",
"model_params": {
"temperature": 0.7,
"max_tokens": 500
},
"model": "qwen3-30b-a3b-deepseek-distill-instruct-2507"
}
```
**Response:**
```
{
"content": [
{
"type": "text",
"text": "Quantum entanglement is a phenomenon where..."
}
]
}
```
---
### `private_analysis`
Analyze content locally without cloud exposure.
**Parameters:**
- `content` (string, required): Text to analyze
- `analysis_type` (enum, required): Type of analysis
- `sentiment`: Detect emotions and sentiment
- `entities`: Extract named entities
- `classification`: Categorize content
- `summary`: Generate summary with key points
- `key_points`: Extract main points and insights
- `privacy_scan`: Detect PII and privacy issues
- `security_audit`: Security vulnerability assessment
- `domain` (enum, optional): Domain context (default: "general")
- `general`, `medical`, `legal`, `financial`, `technical`, `academic`
**Example - Sentiment Analysis:**
```json
{
"content": "The new policy changes are concerning but necessary for growth",
"analysis_type": "sentiment",
"domain": "financial"
}
```
**Response:**
```json
{
"sentiment": "neutral",
"confidence": 0.75,
"emotions": ["concern", "optimism"],
"reasoning": "Mixed sentiment with concern balanced by positive outlook"
}
```
**Example - Entity Extraction:**
```json
{
"content": "Dr. Smith from Mayo Clinic prescribed medication X on January 15, 2024",
"analysis_type": "entities",
"domain": "medical"
}
```
**Response:**
```json
{
"people": ["Dr. Smith"],
"organizations": ["Mayo Clinic"],
"locations": [],
"dates": ["January 15, 2024"],
"domainSpecific": {
"medications": ["medication X"],
"medical_conditions": [],
"procedures": []
}
}
```
**Example - Privacy Scan:**
```json
{
"content": "Contact John Doe at john.doe@email.com or 555-123-4567",
"analysis_type": "privacy_scan"
}
```
**Response:**
```json
{
"issues": [
{
"type": "email",
"severity": "medium",
"description": "Email address detected",
"location": "john.doe@email.com"
},
{
"type": "phone",
"severity": "medium",
"description": "Phone number detected",
"location": "555-123-4567"
}
],
"overallRisk": "medium",
"recommendations": [
"Remove or mask email addresses",
"Remove or mask phone numbers"
]
}
```
---
### `secure_rewrite`
Rewrite text with privacy protection and style transformation.
**Parameters:**
- `content` (string, required): Original text to rewrite
- `style` (string, required): Target style (e.g., "formal", "casual", "professional")
- `privacy_level` (enum, optional): Privacy protection level (default: "moderate")
- `strict`: Maximum privacy, remove all PII
- `moderate`: Balance privacy and readability
- `minimal`: Basic PII protection only
**Example:**
```json
{
"content": "Hey, call me at 555-1234 to discuss the contract",
"style": "professional",
"privacy_level": "strict"
}
```
**Response:**
```
"Please contact me via provided communication channels to discuss the contractual agreement."
```
---
### `code_analysis`
Analyze code locally for security, quality, or documentation.
**Parameters:**
- `code` (string, required): Source code to analyze
- `language` (string, required): Programming language (e.g., "python", "javascript", "java")
- `analysis_focus` (enum, required): Focus area
- `security`: Vulnerability and security issues
- `quality`: Code quality and best practices
- `documentation`: Documentation needs
- `optimization`: Performance optimization
- `bugs`: Potential bugs and errors
**Example:**
```json
{
"code": "def login(username, password):\n query = f\"SELECT * FROM users WHERE name='{username}' AND pwd='{password}'\"\n return db.execute(query)",
"language": "python",
"analysis_focus": "security"
}
```
**Response:**
```json
{
"analysis": "Critical SQL injection vulnerability detected",
"issues": [
{
"line": 2,
"severity": "high",
"description": "SQL Injection vulnerability - user input directly in query",
"suggestion": "Use parameterized queries or ORM to prevent SQL injection"
},
{
"line": 2,
"severity": "high",
"description": "Plaintext password in query",
"suggestion": "Passwords should be hashed before comparison"
}
],
"metrics": {
"linesOfCode": 3,
"functionsFound": 1
},
"recommendations": [
"Implement input validation and sanitization",
"Use parameterized queries to prevent SQL injection",
"Hash passwords before storage and comparison"
]
}
```
---
### `template_completion`
Complete templates with intelligent placeholder filling.
**Parameters:**
- `template` (string, required): Template with `[PLACEHOLDER]` markers
- `context` (string, required): Context information for completion
- `format` (string, optional): Output format requirements
**Example:**
```json
{
"template": "Dear [CUSTOMER_NAME],\n\nThank you for your [REQUEST_TYPE] regarding [SUBJECT].\n\nNext steps:\n1. [STEP_1]\n2. [STEP_2]\n3. [STEP_3]\n\nBest regards,\n[AGENT_NAME]",
"context": "Customer John Smith requested a premium account upgrade for advanced analytics features"
}
```
**Response:**
```
"Dear John Smith,
Thank you for your upgrade request regarding premium analytics features.
Next steps:
1. Verify your payment method
2. Explore your new advanced features
3. Contact us if you have any questions
Best regards,
Customer Service Team"
```
---
### `set_default_model`
Change the default model for the current session.
**Parameters:**
- `model` (string, required): Model name from available models
**Example:**
```json
{
"model": "qwen3-30b-a3b-deepseek-distill-instruct-2507"
}
```
**Response:**
```json
{
"success": true,
"defaultModel": "qwen3-30b-a3b-deepseek-distill-instruct-2507",
"previousDefault": "openai/gpt-oss-20b",
"message": "Default model set to \"qwen3-30b-a3b-deepseek-distill-instruct-2507\""
}
```
**Error Response (invalid model):**
```json
{
"error": "Model \"invalid-model\" not found",
"availableModels": [
"openai/gpt-oss-20b",
"qwen3-30b-a3b-deepseek-distill-instruct-2507"
],
"currentDefault": "openai/gpt-oss-20b",
"message": "Model \"invalid-model\" is not available. Choose from the available models listed above."
}
```
---
## Resources
### `local://models`
List all models loaded in LM Studio with current default.
**Response:**
```json
{
"models": [
"openai/gpt-oss-20b",
"qwen3-30b-a3b-deepseek-distill-instruct-2507",
"text-embedding-nomic-embed-text-v1.5"
],
"count": 3,
"defaultModel": "openai/gpt-oss-20b"
}
```
---
### `local://status`
Check if LM Studio server is online/offline.
**Response:**
```json
{
"status": "online",
"timestamp": "2025-01-05T02:22:09.412Z"
}
```
---
### `local://config`
View server capabilities, domains, privacy levels, and analysis types.
**Response:**
```json
{
"server": "local-llm-mcp-server",
"version": "1.0.0",
"capabilities": [
"reasoning",
"analysis",
"rewriting",
"code_analysis",
"templates"
],
"privacy_levels": ["strict", "moderate", "minimal"],
"supported_domains": [
"general",
"medical",
"legal",
"financial",
"technical",
"academic"
],
"analysis_types": [
"sentiment",
"entities",
"classification",
"summary",
"key_points",
"privacy_scan",
"security_audit"
]
}
```
---
### `local://capabilities`
Comprehensive documentation including tool examples and workflow guide.
**Response:** (Abbreviated example)
```json
{
"modelManagement": {
"availableModels": ["openai/gpt-oss-20b", "qwen3-30b..."],
"defaultModel": "openai/gpt-oss-20b",
"howToUse": {
"readModels": "Read resource: local://models",
"setDefault": "Use tool: set_default_model with {\"model\": \"model-name\"}",
"useSpecificModel": "Add \"model\" parameter to any tool call",
"useDefaultModel": "Omit \"model\" parameter to use the default model"
}
},
"tools": {
"local_reasoning": {
"purpose": "General-purpose reasoning and question answering",
"parameters": {...},
"example": "{\"prompt\": \"Explain quantum computing\", \"model\": \"openai/gpt-oss-20b\"}"
}
// ... other tools
},
"workflow": {
"step1": "Read local://models to see available models",
"step2": "Optionally set a default model using set_default_model tool",
"step3": "Use any tool - it will use default model unless you specify \"model\" parameter",
"step4": "Override default by passing \"model\" parameter in any tool call"
}
}
```
---
## Prompts
Pre-built prompt templates for common tasks.
### Available Prompts
1. **`privacy-analysis`** - Analyze content for privacy issues
2. **`secure-rewrite`** - Rewrite with privacy protection
3. **`document-analysis`** - Comprehensive document analysis
4. **`risk-assessment`** - Risk analysis with mitigation
5. **`code-security-review`** - Security-focused code review
6. **`code-optimization`** - Code optimization suggestions
7. **`meeting-summary`** - Meeting summaries with action items
8. **`email-draft`** - Professional email drafting
9. **`research-synthesis`** - Synthesize research sources
10. **`literature-review`** - Systematic literature review
11. **`content-adaptation`** - Adapt content for different audiences
12. **`technical-documentation`** - Generate technical docs
---
## Model Management
### Workflow
1. **Discover Models**
```
Read resource: local://models
```
2. **View Current Default**
```json
// In response from local://models
{
"defaultModel": "openai/gpt-oss-20b"
}
```
3. **Change Default**
```json
{
"tool": "set_default_model",
"arguments": {
"model": "qwen3-30b-a3b-deepseek-distill-instruct-2507"
}
}
```
4. **Use Default Model**
```json
{
"tool": "local_reasoning",
"arguments": {
"prompt": "Your question here"
// No model specified - uses default
}
}
```
5. **Override for Specific Request**
```json
{
"tool": "local_reasoning",
"arguments": {
"prompt": "Your question here",
"model": "openai/gpt-oss-20b" // Overrides default
}
}
```
---
## Error Handling
### Common Errors
**Model Not Found:**
```json
{
"error": "Model \"invalid-model\" not found",
"availableModels": ["model1", "model2"],
"currentDefault": "model1",
"message": "Model \"invalid-model\" is not available..."
}
```
**No Default Model Set:**
```
"Error: No model specified and no default model set. Use setDefaultModel() or pass a model parameter."
```
**LM Studio Connection Failed:**
```
"Error: Failed to generate response: Failed to connect to LM Studio"
```
**Timeout:**
```
"Error: Request timed out after 600s. Performance suggestions: Reduce max_tokens to 1000 or less, Try using a faster/smaller model..."
```
### Best Practices
1. **Always check `local://models` first** to see available models
2. **Set a default model** for your session to avoid specifying it every time
3. **Use appropriate model parameters** based on your task
4. **Handle errors gracefully** - check for `isError: true` in responses
5. **Monitor performance** - adjust `max_tokens` and `temperature` as needed
