@jaypie/mcp 0.7.4 → 0.7.7

package/skills/tools-llm.md

@@ -0,0 +1,98 @@
+---
+description: LLM MCP tool for debugging provider responses
+related: llm, tools
+---
+
+# LLM MCP Tool
+
+Debug and inspect LLM provider responses. Useful for understanding how providers format responses and troubleshooting API integrations.
+
+## Usage
+
+```
+llm()                                    # Show help
+llm("command", { ...params })            # Execute a command
+```
+
+## Commands
+
+| Command | Description |
+|---------|-------------|
+| `list_providers` | List available LLM providers and their status |
+| `debug_call` | Make a debug call to an LLM provider |
+
+## List Providers
+
+Check which providers are configured:
+
+```
+llm("list_providers")
+```
+
+Returns provider availability based on environment variables.
+
+## Debug Call
+
+Make a test call to inspect the raw response:
+
+```
+llm("debug_call", { provider: "openai", message: "Hello, world!" })
+llm("debug_call", { provider: "anthropic", message: "Hello, world!" })
+llm("debug_call", { provider: "openai", model: "o3-mini", message: "What is 15 * 17?" })
+```
+
+### Parameters
+
+| Parameter | Required | Description |
+|-----------|----------|-------------|
+| `provider` | Yes | Provider name: `openai`, `anthropic`, `gemini`, `openrouter` |
+| `message` | Yes | Message to send |
+| `model` | No | Specific model to use |
+
+### Response Fields
+
+| Field | Description |
+|-------|-------------|
+| `content` | The response text |
+| `reasoning` | Extracted reasoning/thinking content (if available) |
+| `reasoningTokens` | Count of reasoning tokens used |
+| `history` | Full conversation history |
+| `rawResponses` | Raw API responses for debugging |
+| `usage` | Token usage statistics |
+
+## Environment Variables
+
+| Variable | Description |
+|----------|-------------|
+| `OPENAI_API_KEY` | OpenAI API key |
+| `ANTHROPIC_API_KEY` | Anthropic API key |
+| `GOOGLE_API_KEY` | Google/Gemini API key |
+| `OPENROUTER_API_KEY` | OpenRouter API key |
+
+## Supported Providers
+
+| Provider | Models |
+|----------|--------|
+| `openai` | gpt-4o, gpt-4o-mini, o1, o3-mini, etc. |
+| `anthropic` | claude-sonnet-4-20250514, claude-opus-4-20250514, etc. |
+| `gemini` | gemini-2.0-flash, gemini-1.5-pro, etc. |
+| `openrouter` | Access to multiple providers |
+
+## Common Patterns
+
+### Compare Provider Responses
+
+```
+llm("debug_call", { provider: "openai", message: "Explain recursion briefly" })
+llm("debug_call", { provider: "anthropic", message: "Explain recursion briefly" })
+```
+
+### Test Reasoning Models
+
+```
+llm("debug_call", { provider: "openai", model: "o3-mini", message: "Solve: If 3x + 5 = 14, what is x?" })
+```
+
+### Check Token Usage
+
+Use `debug_call` to inspect the `usage` field for token consumption analysis.

package/skills/tools.md

@@ -1,6 +1,6 @@
 ---
 description: Available MCP tools reference
-related: tools-aws, tools-dynamodb, tools-datadog, debugging
+related: tools-aws, tools-datadog, tools-dynamodb, tools-llm, debugging
 ---
 
 # MCP Tools Reference
@@ -79,6 +79,10 @@ datadog("monitors", { status: ["Alert", "Warn"] })
 
 ## LLM Tool
 
+Debug and inspect LLM provider responses.
+
+See **tools-llm** for complete documentation.
+
 ```
 llm()                   # Show help
 llm("list_providers")   # List available LLM providers
@@ -103,3 +107,9 @@ llm("debug_call", { provider: "openai", message: "Hello" })
 - `DD_SERVICE` - Default service filter
 - `DD_SOURCE` - Default log source
 
+### LLM Tools
+- `OPENAI_API_KEY` - OpenAI API key
+- `ANTHROPIC_API_KEY` - Anthropic API key
+- `GOOGLE_API_KEY` - Google/Gemini API key
+- `OPENROUTER_API_KEY` - OpenRouter API key
+