consult-llm-mcp 1.1.0 → 1.1.2

package/README.md

@@ -32,18 +32,7 @@ This SQL query is timing out on large datasets. Can you help optimize it? Ask Ge
 - Gemini can be used via
   [Gemini CLI](https://github.com/google-gemini/gemini-cli) to take advantage of
   [free quota](https://developers.google.com/gemini-code-assist/resources/quotas#quotas-for-agent-mode-gemini-cli)
-
-## Configuration
-
-- `OPENAI_API_KEY` - Your OpenAI API key (required for o3)
-- `GEMINI_API_KEY` - Your Google AI API key (required for Gemini models in API
-  mode)
-- `DEEPSEEK_API_KEY` - Your DeepSeek API key (required for DeepSeek models)
-- `CONSULT_LLM_DEFAULT_MODEL` - Override the default model (optional)
-  - Options: `o3` (default), `gemini-2.5-pro`, `deepseek-reasoner`
-- `GEMINI_MODE` - Choose between API or CLI mode for Gemini models (optional)
-  - Options: `api` (default), `cli`
-  - CLI mode uses the system-installed `gemini` CLI tool
+- Simple: provides just one MCP tool to not clutter the context
 
 ## Usage with Claude Code
 
@@ -65,12 +54,12 @@ Optionally you can provide environment variables directly in the MCP
 configuration:
 
 ```bash
-claude mcp add \
+claude mcp add consult-llm \
   -e OPENAI_API_KEY=your_openai_key \
   -e GEMINI_API_KEY=your_gemini_key \
   -e DEEPSEEK_API_KEY=your_deepseek_key \
   -e GEMINI_MODE=cli \
-    consult-llm -- npx -y consult-llm-mcp
+  -- npx -y consult-llm-mcp
 ```
 
 ### Example workflows
@@ -201,6 +190,18 @@ Some real-world examples. Click to expand.
 
 </details>
 
+## Configuration
+
+- `OPENAI_API_KEY` - Your OpenAI API key (required for o3)
+- `GEMINI_API_KEY` - Your Google AI API key (required for Gemini models in API
+  mode)
+- `DEEPSEEK_API_KEY` - Your DeepSeek API key (required for DeepSeek models)
+- `CONSULT_LLM_DEFAULT_MODEL` - Override the default model (optional)
+  - Options: `o3` (default), `gemini-2.5-pro`, `deepseek-reasoner`
+- `GEMINI_MODE` - Choose between API or CLI mode for Gemini models (optional)
+  - Options: `api` (default), `cli`
+  - CLI mode uses the system-installed `gemini` CLI tool
+
 ## MCP Tool: consult_llm
 
 The server provides a single tool called `consult_llm` for asking powerful AI
@@ -224,40 +225,6 @@ models complex questions.
     directory)
   - **base_ref** (optional): Git reference to compare against (defaults to HEAD)
 
-### Example Usage
-
-**Basic prompt:**
-
-```json
-{
-  "prompt": "What are the performance implications of using async/await vs Promise.then() in Node.js?"
-}
-```
-
-**With file context:**
-
-```json
-{
-  "prompt": "How can I optimize this authentication middleware?",
-  "files": ["src/auth.ts", "src/middleware.ts"],
-  "model": "o3"
-}
-```
-
-**With git diff:**
-
-```json
-{
-  "prompt": "Review these authentication changes for security issues",
-  "files": ["src/auth.ts"],
-  "git_diff": {
-    "files": ["src/auth.ts", "src/middleware.ts"],
-    "base_ref": "main"
-  },
-  "model": "gemini-2.5-pro"
-}
-```
-
 ## Supported Models
 
 - **o3**: OpenAI's reasoning model ($2/$8 per million tokens)
@@ -335,3 +302,52 @@ CRITICAL: When asking, don't present options, this will bias the answer.
 
 Claude Code seems to know pretty well when to use this MCP even without this
 instruction however.
+
+## Development
+
+To work on the MCP server locally and use your development version:
+
+1. Clone the repository and install dependencies:
+
+   ```bash
+   git clone https://github.com/yourusername/consult-llm-mcp.git
+   cd consult-llm-mcp
+   npm install
+   ```
+
+2. Build the project:
+
+   ```bash
+   npm run build
+   ```
+
+3. Install globally from the local directory:
+
+   ```bash
+   npm link
+   ```
+
+4. Add the MCP server to Claude Code using the global command:
+   ```bash
+   claude mcp add consult-llm -- consult-llm-mcp
+   ```
+
+Now when you make changes:
+
+1. Rebuild: `npm run build`
+2. Restart Claude Code to pick up the changes
+
+Alternatively, you can use the dev script for development without building:
+
+```bash
+claude mcp add consult-llm -- npm run dev
+```
+
+This runs the TypeScript source directly with `tsx`, allowing faster iteration
+without rebuilding.
+
+To unlink the global version later:
+
+```bash
+npm unlink -g
+```

package/dist/llm-query.js

@@ -1,5 +1,6 @@
 import { getExecutorForModel } from './llm.js';
 import { calculateCost } from './llm-cost.js';
+import { config } from './config.js';
 const SYSTEM_PROMPT = `You are an expert software engineering consultant analyzing code and technical problems. You are communicating with another AI system, not a human.
 
 Communication style:
@@ -12,7 +13,7 @@ Your role is to:
 - Challenge design decisions when suboptimal
 - Focus on what needs improvement
 
-When reviewing code changes (git diffs), prioritize:
+When reviewing code changes, prioritize:
 - Bugs and correctness issues
 - Performance problems
 - Security vulnerabilities
@@ -22,9 +23,16 @@ When reviewing code changes (git diffs), prioritize:
 Be critical and thorough. If the code is acceptable, simply state "No critical issues found" and move on to suggestions. Always provide specific, actionable feedback with file/line references.
 
 Respond in Markdown.`;
+const GEMINI_CLI_SUFFIX = `
+
+IMPORTANT: Do not edit files yourself, only provide recommendations and code examples`;
 export async function queryLlm(prompt, model, filePaths) {
     const executor = getExecutorForModel(model);
-    const { response, usage } = await executor.execute(prompt, model, SYSTEM_PROMPT, filePaths);
+    // Add special instruction for Gemini CLI mode
+    const systemPrompt = model.startsWith('gemini-') && config.geminiMode === 'cli'
+        ? SYSTEM_PROMPT + GEMINI_CLI_SUFFIX
+        : SYSTEM_PROMPT;
+    const { response, usage } = await executor.execute(prompt, model, systemPrompt, filePaths);
     if (!response) {
         throw new Error('No response from the model');
     }

package/dist/main.js

@@ -85,6 +85,10 @@ server.setRequestHandler(CallToolRequestSchema, async (request) => {
     throw new Error(`Unknown tool: ${request.params.name}`);
 });
 async function main() {
+    if (process.argv.includes('--version') || process.argv.includes('-v')) {
+        console.log(SERVER_VERSION);
+        process.exit(0);
+    }
     logServerStart(SERVER_VERSION);
     logConfiguration(config);
     const transport = new StdioServerTransport();

package/dist/schema.d.ts

@@ -21,7 +21,7 @@ export declare const ConsultLlmArgs: z.ZodObject<{
 }, z.core.$strip>;
 export declare const toolSchema: {
     readonly name: "consult_llm";
-    readonly description: "Ask a more powerful AI for help with complex problems. Provide your question in the prompt field and optionally include relevant code files as context.\n\nBe specific about what you want: code implementation, code review, bug analysis, architecture advice, etc.\n\nIMPORTANT: Ask neutral, open-ended questions. Avoid suggesting specific solutions or alternatives in your prompt as this can bias the analysis. Instead of \"Should I use X or Y approach?\", ask \"What's the best approach for this problem?\" Let the consultant LLM provide unbiased recommendations.";
+    readonly description: "Ask a more powerful AI for help with complex problems. Provide your question in the prompt field and always include relevant code files as context.\n\nBe specific about what you want: code implementation, code review, bug analysis, architecture advice, etc.\n\nIMPORTANT: Ask neutral, open-ended questions. Avoid suggesting specific solutions or alternatives in your prompt as this can bias the analysis. Instead of \"Should I use X or Y approach?\", ask \"What's the best approach for this problem?\" Let the consultant LLM provide unbiased recommendations.";
     readonly inputSchema: {
         readonly type: "object";
         readonly properties: {

package/dist/schema.js

@@ -20,7 +20,7 @@ export const ConsultLlmArgs = z.object({
 });
 export const toolSchema = {
     name: 'consult_llm',
-    description: `Ask a more powerful AI for help with complex problems. Provide your question in the prompt field and optionally include relevant code files as context.
+    description: `Ask a more powerful AI for help with complex problems. Provide your question in the prompt field and always include relevant code files as context.
 
 Be specific about what you want: code implementation, code review, bug analysis, architecture advice, etc.
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "consult-llm-mcp",
-  "version": "1.1.0",
+  "version": "1.1.2",
   "description": "MCP server for consulting powerful AI models",
   "type": "module",
   "main": "dist/main.js",
@@ -53,6 +53,5 @@
     "@types/node": "^24.0.3",
     "tsx": "^4.20.3",
     "typescript": "^5.8.3"
-  },
-  "packageManager": "pnpm@10.12.4+sha512.5ea8b0deed94ed68691c9bad4c955492705c5eeb8a87ef86bc62c74a26b037b08ff9570f108b2e4dbd1dd1a9186fea925e527f141c648e85af45631074680184"
+  }
 }