consult-llm-mcp 2.4.0 → 2.4.2

package/README.md

@@ -1,7 +1,7 @@
 # Consult LLM MCP
 
 An MCP server that lets Claude Code consult stronger AI models (GPT-5.2, Gemini
-3.0 Pro, DeepSeek Reasoner) when Sonnet has you running in circles and you need
+3.1 Pro, DeepSeek Reasoner) when Sonnet has you running in circles and you need
 to bring in the heavy artillery. Supports multi-turn conversations.
 
 ```
@@ -27,7 +27,7 @@ to bring in the heavy artillery. Supports multi-turn conversations.
 
 ## Features
 
-- Query powerful AI models (GPT-5.2, Gemini 3.0 Pro, DeepSeek Reasoner) with
+- Query powerful AI models (GPT-5.2, Gemini 3.1 Pro, DeepSeek Reasoner) with
   relevant files as context
 - Direct queries with optional file context
 - Include git changes for code review and analysis
@@ -333,7 +333,8 @@ context, but it helps.
 #### Gemini CLI
 
 Use Gemini's local CLI to take advantage of Google's
-[free quota](https://developers.google.com/gemini-code-assist/resources/quotas#quotas-for-agent-mode-gemini-cli).
+[free quota](https://developers.google.com/gemini-code-assist/resources/quotas#quotas-for-agent-mode-gemini-cli)
+or use your Google AI Pro subscription.
 
 **Requirements:**
 
@@ -456,7 +457,8 @@ See the "Using web mode..." example above for a concrete transcript.
 - `DEEPSEEK_API_KEY` - Your DeepSeek API key (required for DeepSeek models)
 - `CONSULT_LLM_DEFAULT_MODEL` - Override the default model (optional)
   - Options: `gpt-5.2` (default), `gemini-2.5-pro`, `gemini-3-pro-preview`,
-    `deepseek-reasoner`, `gpt-5.3-codex`, `gpt-5.2-codex`
+    `gemini-3.1-pro-preview`, `deepseek-reasoner`, `gpt-5.3-codex`,
+    `gpt-5.2-codex`
 - `GEMINI_BACKEND` - Backend for Gemini models (optional)
   - Options: `api` (default), `gemini-cli`, `cursor-cli`
 - `OPENAI_BACKEND` - Backend for OpenAI models (optional)
@@ -554,7 +556,8 @@ models complex questions.
 
 - **model** (optional): LLM model to use
   - Options: `gpt-5.2` (default), `gemini-2.5-pro`, `gemini-3-pro-preview`,
-    `deepseek-reasoner`, `gpt-5.3-codex`, `gpt-5.2-codex`
+    `gemini-3.1-pro-preview`, `deepseek-reasoner`, `gpt-5.3-codex`,
+    `gpt-5.2-codex`
 
 - **task_mode** (optional): Controls the system prompt persona. The calling LLM
   should choose based on the task:
@@ -589,6 +592,8 @@ models complex questions.
 - **gemini-2.5-pro**: Google's Gemini 2.5 Pro ($1.25/$10 per million tokens)
 - **gemini-3-pro-preview**: Google's Gemini 3 Pro Preview ($2/$12 per million
   tokens for prompts ≤200k tokens, $4/$18 for prompts >200k tokens)
+- **gemini-3.1-pro-preview**: Google's Gemini 3.1 Pro Preview ($2/$12 per
+  million tokens for prompts ≤200k tokens, $4/$18 for prompts >200k tokens)
 - **deepseek-reasoner**: DeepSeek's reasoning model ($0.55/$2.19 per million
   tokens)
 - **gpt-5.2**: OpenAI's latest GPT model

package/dist/config.d.ts

@@ -1,4 +1,12 @@
 import { z } from 'zod/v4';
+export interface ProviderAvailability {
+    geminiApiKey?: string;
+    geminiBackend: string;
+    openaiApiKey?: string;
+    openaiBackend: string;
+    deepseekApiKey?: string;
+}
+export declare function filterByAvailability(models: string[], providers: ProviderAvailability): string[];
 /** Build the final model catalog from built-in + extra + allowlist filtering. */
 export declare function buildModelCatalog(builtinModels: readonly string[], extraModelsRaw?: string, allowedModelsRaw?: string): string[];
 export declare const SupportedChatModel: z.ZodEnum<{

package/dist/config.js

@@ -1,6 +1,21 @@
 import { z } from 'zod/v4';
 import { ALL_MODELS } from './models.js';
 import { logToFile } from './logger.js';
+export function filterByAvailability(models, providers) {
+    return models.filter((model) => {
+        if (model.startsWith('gemini-')) {
+            return providers.geminiBackend !== 'api' || !!providers.geminiApiKey;
+        }
+        if (model.startsWith('gpt-')) {
+            return providers.openaiBackend !== 'api' || !!providers.openaiApiKey;
+        }
+        if (model.startsWith('deepseek-')) {
+            return !!providers.deepseekApiKey;
+        }
+        // Unknown prefix (user-added extra models) — always include
+        return true;
+    });
+}
 /** Build the final model catalog from built-in + extra + allowlist filtering. */
 export function buildModelCatalog(builtinModels, extraModelsRaw, allowedModelsRaw) {
     const extraModels = extraModelsRaw
@@ -23,9 +38,20 @@ export function buildModelCatalog(builtinModels, extraModelsRaw, allowedModelsRa
         ? allAvailable.filter((m) => allowedModels.includes(m))
         : allAvailable;
 }
-const enabledModels = buildModelCatalog(ALL_MODELS, process.env.CONSULT_LLM_EXTRA_MODELS, process.env.CONSULT_LLM_ALLOWED_MODELS);
+// Resolve backends early (needed for availability filtering)
+const resolvedGeminiBackend = migrateBackendEnv(process.env.GEMINI_BACKEND, process.env.GEMINI_MODE, 'gemini-cli', 'GEMINI_MODE', 'GEMINI_BACKEND');
+const resolvedOpenaiBackend = migrateBackendEnv(process.env.OPENAI_BACKEND, process.env.OPENAI_MODE, 'codex-cli', 'OPENAI_MODE', 'OPENAI_BACKEND');
+// Build catalog, then filter to only available providers
+const catalogModels = buildModelCatalog(ALL_MODELS, process.env.CONSULT_LLM_EXTRA_MODELS, process.env.CONSULT_LLM_ALLOWED_MODELS);
+const enabledModels = filterByAvailability(catalogModels, {
+    geminiApiKey: process.env.GEMINI_API_KEY,
+    geminiBackend: resolvedGeminiBackend ?? 'api',
+    openaiApiKey: process.env.OPENAI_API_KEY,
+    openaiBackend: resolvedOpenaiBackend ?? 'api',
+    deepseekApiKey: process.env.DEEPSEEK_API_KEY,
+});
 if (enabledModels.length === 0) {
-    const msg = 'Invalid environment variables:\n  CONSULT_LLM_ALLOWED_MODELS: No valid models enabled.';
+    const msg = 'Invalid environment variables:\n  No models available. Set API keys or configure CLI backends.';
     logToFile(`FATAL ERROR:\n${msg}`);
     console.error(`❌ ${msg}`);
     process.exit(1);
@@ -62,8 +88,8 @@ const parsedConfig = Config.safeParse({
     geminiApiKey: process.env.GEMINI_API_KEY,
     deepseekApiKey: process.env.DEEPSEEK_API_KEY,
     defaultModel: process.env.CONSULT_LLM_DEFAULT_MODEL,
-    geminiBackend: migrateBackendEnv(process.env.GEMINI_BACKEND, process.env.GEMINI_MODE, 'gemini-cli', 'GEMINI_MODE', 'GEMINI_BACKEND'),
-    openaiBackend: migrateBackendEnv(process.env.OPENAI_BACKEND, process.env.OPENAI_MODE, 'codex-cli', 'OPENAI_MODE', 'OPENAI_BACKEND'),
+    geminiBackend: resolvedGeminiBackend,
+    openaiBackend: resolvedOpenaiBackend,
     codexReasoningEffort: process.env.CODEX_REASONING_EFFORT,
     systemPromptPath: process.env.CONSULT_LLM_SYSTEM_PROMPT_PATH,
 });

package/dist/config.test.js

@@ -1,5 +1,5 @@
 import { describe, it, expect, vi } from 'vitest';
-import { migrateBackendEnv, buildModelCatalog } from './config.js';
+import { migrateBackendEnv, buildModelCatalog, filterByAvailability, } from './config.js';
 import { ALL_MODELS } from './models.js';
 vi.mock('./logger.js', () => ({ logToFile: vi.fn() }));
 describe('migrateBackendEnv', () => {
@@ -19,6 +19,94 @@ describe('migrateBackendEnv', () => {
         expect(migrateBackendEnv(undefined, 'cli', 'codex-cli', 'OPENAI_MODE', 'OPENAI_BACKEND')).toBe('codex-cli');
     });
 });
+describe('filterByAvailability', () => {
+    const allModels = ['gemini-2.5-pro', 'gpt-5.2', 'deepseek-reasoner'];
+    it('includes gemini models when using gemini-cli backend', () => {
+        const result = filterByAvailability(allModels, {
+            geminiBackend: 'gemini-cli',
+            openaiBackend: 'api',
+        });
+        expect(result).toContain('gemini-2.5-pro');
+        expect(result).not.toContain('gpt-5.2');
+    });
+    it('includes gemini models when using cursor-cli backend', () => {
+        const result = filterByAvailability(allModels, {
+            geminiBackend: 'cursor-cli',
+            openaiBackend: 'api',
+        });
+        expect(result).toContain('gemini-2.5-pro');
+    });
+    it('includes gemini models when API key is set', () => {
+        const result = filterByAvailability(allModels, {
+            geminiApiKey: 'key',
+            geminiBackend: 'api',
+            openaiBackend: 'api',
+        });
+        expect(result).toContain('gemini-2.5-pro');
+    });
+    it('excludes gemini models when backend is api and no key', () => {
+        const result = filterByAvailability(allModels, {
+            geminiBackend: 'api',
+            openaiBackend: 'api',
+        });
+        expect(result).not.toContain('gemini-2.5-pro');
+    });
+    it('includes gpt models when using codex-cli backend', () => {
+        const result = filterByAvailability(allModels, {
+            geminiBackend: 'api',
+            openaiBackend: 'codex-cli',
+        });
+        expect(result).toContain('gpt-5.2');
+        expect(result).not.toContain('gemini-2.5-pro');
+    });
+    it('includes gpt models when using cursor-cli backend', () => {
+        const result = filterByAvailability(allModels, {
+            geminiBackend: 'api',
+            openaiBackend: 'cursor-cli',
+        });
+        expect(result).toContain('gpt-5.2');
+    });
+    it('includes gpt models when API key is set', () => {
+        const result = filterByAvailability(allModels, {
+            geminiBackend: 'api',
+            openaiApiKey: 'key',
+            openaiBackend: 'api',
+        });
+        expect(result).toContain('gpt-5.2');
+    });
+    it('excludes gpt models when backend is api and no key', () => {
+        const result = filterByAvailability(allModels, {
+            geminiBackend: 'api',
+            openaiBackend: 'api',
+        });
+        expect(result).not.toContain('gpt-5.2');
+    });
+    it('includes deepseek models only when API key is set', () => {
+        expect(filterByAvailability(allModels, {
+            geminiBackend: 'api',
+            openaiBackend: 'api',
+            deepseekApiKey: 'key',
+        })).toContain('deepseek-reasoner');
+        expect(filterByAvailability(allModels, {
+            geminiBackend: 'api',
+            openaiBackend: 'api',
+        })).not.toContain('deepseek-reasoner');
+    });
+    it('includes unknown-prefix models (user extras)', () => {
+        const result = filterByAvailability([...allModels, 'grok-3'], {
+            geminiBackend: 'api',
+            openaiBackend: 'api',
+        });
+        expect(result).toContain('grok-3');
+    });
+    it('returns empty array when no providers are available', () => {
+        const result = filterByAvailability(allModels, {
+            geminiBackend: 'api',
+            openaiBackend: 'api',
+        });
+        expect(result).toEqual([]);
+    });
+});
 describe('buildModelCatalog', () => {
     it('returns all built-in models when no env vars are set', () => {
         const result = buildModelCatalog(ALL_MODELS);

package/dist/executors/cli-runner.d.ts

@@ -0,0 +1,17 @@
+export interface CliResult {
+    stdout: string;
+    stderr: string;
+    code: number | null;
+    duration: number;
+}
+/**
+ * Spawn a CLI process and collect its output.
+ *
+ * Two modes:
+ * - **Buffered** (default): collects all stdout into `result.stdout`
+ * - **Streaming**: pass `onLine` to process each stdout line as it arrives;
+ *   `result.stdout` will be empty
+ */
+export declare function runCli(command: string, args: string[], options?: {
+    onLine?: (line: string) => void;
+}): Promise<CliResult>;

package/dist/executors/cli-runner.js

@@ -0,0 +1,57 @@
+import { createInterface } from 'node:readline';
+import { spawn } from 'node:child_process';
+import { logCliDebug } from '../logger.js';
+/**
+ * Spawn a CLI process and collect its output.
+ *
+ * Two modes:
+ * - **Buffered** (default): collects all stdout into `result.stdout`
+ * - **Streaming**: pass `onLine` to process each stdout line as it arrives;
+ *   `result.stdout` will be empty
+ */
+export function runCli(command, args, options) {
+    return new Promise((resolve, reject) => {
+        try {
+            logCliDebug(`Spawning ${command} CLI`, {
+                promptLength: args[args.length - 1]?.length,
+                args,
+            });
+            const child = spawn(command, args, {
+                shell: false,
+                stdio: ['ignore', 'pipe', 'pipe'],
+            });
+            let stdout = '';
+            let stderr = '';
+            const startTime = Date.now();
+            child.on('spawn', () => logCliDebug(`${command} CLI process spawned successfully`));
+            if (options?.onLine) {
+                const rl = createInterface({ input: child.stdout, terminal: false });
+                rl.on('line', options.onLine);
+            }
+            else {
+                child.stdout.setEncoding('utf8');
+                child.stdout.on('data', (chunk) => (stdout += chunk));
+            }
+            child.stderr.on('data', (data) => (stderr += data.toString()));
+            child.on('close', (code) => {
+                const duration = Date.now() - startTime;
+                logCliDebug(`${command} CLI process closed`, {
+                    code,
+                    duration: `${duration}ms`,
+                    stdoutLength: stdout.length,
+                    stderrLength: stderr.length,
+                });
+                resolve({ stdout, stderr, code, duration });
+            });
+            child.on('error', (err) => {
+                logCliDebug(`Failed to spawn ${command} CLI`, {
+                    error: err.message,
+                });
+                reject(new Error(`Failed to spawn ${command} CLI. Is it installed and in PATH? Error: ${err.message}`));
+            });
+        }
+        catch (err) {
+            reject(new Error(`Synchronous error while trying to spawn ${command}: ${err instanceof Error ? err.message : String(err)}`));
+        }
+    });
+}

package/dist/executors/codex-cli.js

@@ -1,7 +1,6 @@
-import { spawn } from 'node:child_process';
 import { relative } from 'node:path';
 import { config } from '../config.js';
-import { logCliDebug } from '../logger.js';
+import { runCli } from './cli-runner.js';
 export function parseCodexJsonl(output) {
     let threadId;
     const messages = [];
@@ -61,57 +60,21 @@ export function createCodexExecutor() {
                 }
                 args.push('-m', model, fullPrompt);
             }
-            return new Promise((resolve, reject) => {
-                try {
-                    logCliDebug('Spawning codex CLI', {
-                        model,
-                        promptLength: fullPrompt.length,
-                        threadId,
-                        args,
-                    });
-                    const child = spawn('codex', args, {
-                        shell: false,
-                        stdio: ['ignore', 'pipe', 'pipe'],
-                    });
-                    let stdout = '';
-                    let stderr = '';
-                    const startTime = Date.now();
-                    child.on('spawn', () => logCliDebug('codex CLI process spawned successfully'));
-                    child.stdout.on('data', (data) => (stdout += data.toString()));
-                    child.stderr.on('data', (data) => (stderr += data.toString()));
-                    child.on('close', (code) => {
-                        const duration = Date.now() - startTime;
-                        logCliDebug('codex CLI process closed', {
-                            code,
-                            duration: `${duration}ms`,
-                            stdoutLength: stdout.length,
-                            stderrLength: stderr.length,
-                        });
-                        if (code === 0) {
-                            const parsed = parseCodexJsonl(stdout);
-                            if (!parsed.response) {
-                                reject(new Error('No agent_message found in Codex JSONL output'));
-                                return;
-                            }
-                            resolve({
-                                response: parsed.response,
-                                usage: null,
-                                threadId: parsed.threadId,
-                            });
-                        }
-                        else {
-                            reject(new Error(`Codex CLI exited with code ${code ?? -1}. Error: ${stderr.trim()}`));
-                        }
-                    });
-                    child.on('error', (err) => {
-                        logCliDebug('Failed to spawn codex CLI', { error: err.message });
-                        reject(new Error(`Failed to spawn codex CLI. Is it installed and in PATH? Error: ${err.message}`));
-                    });
+            const { stdout, stderr, code } = await runCli('codex', args);
+            if (code === 0) {
+                const parsed = parseCodexJsonl(stdout);
+                if (!parsed.response) {
+                    throw new Error('No agent_message found in Codex JSONL output');
                 }
-                catch (err) {
-                    reject(new Error(`Synchronous error while trying to spawn codex: ${err instanceof Error ? err.message : String(err)}`));
-                }
-            });
+                return {
+                    response: parsed.response,
+                    usage: null,
+                    threadId: parsed.threadId,
+                };
+            }
+            else {
+                throw new Error(`Codex CLI exited with code ${code ?? -1}. Error: ${stderr.trim()}`);
+            }
         },
     };
 }

package/dist/executors/gemini-cli.js

@@ -1,6 +1,6 @@
-import { spawn } from 'node:child_process';
 import { relative } from 'node:path';
 import { logCliDebug } from '../logger.js';
+import { runCli } from './cli-runner.js';
 export function parseGeminiJson(output) {
     const parsed = JSON.parse(output);
     return {
@@ -33,70 +33,35 @@ export function createGeminiExecutor() {
                 args.push('-r', threadId);
             }
             args.push('-p', message);
-            return new Promise((resolve, reject) => {
+            const { stdout, stderr, code } = await runCli('gemini', args);
+            if (code === 0) {
                 try {
-                    logCliDebug('Spawning gemini CLI', {
-                        model,
-                        promptLength: message.length,
-                        threadId,
-                        args,
-                    });
-                    const child = spawn('gemini', args, {
-                        shell: false,
-                        stdio: ['ignore', 'pipe', 'pipe'],
-                    });
-                    let stdout = '';
-                    let stderr = '';
-                    const startTime = Date.now();
-                    child.on('spawn', () => logCliDebug('gemini CLI process spawned successfully'));
-                    child.stdout.on('data', (data) => (stdout += data.toString()));
-                    child.stderr.on('data', (data) => (stderr += data.toString()));
-                    child.on('close', (code) => {
-                        const duration = Date.now() - startTime;
-                        logCliDebug('gemini CLI process closed', {
-                            code,
-                            duration: `${duration}ms`,
-                            stdoutLength: stdout.length,
-                            stderrLength: stderr.length,
-                        });
-                        if (code === 0) {
-                            try {
-                                const parsed = parseGeminiJson(stdout);
-                                if (!parsed.response) {
-                                    reject(new Error('No response found in Gemini JSON output'));
-                                    return;
-                                }
-                                resolve({
-                                    response: parsed.response,
-                                    usage: null,
-                                    threadId: parsed.sessionId,
-                                });
-                            }
-                            catch {
-                                logCliDebug('Failed to parse Gemini JSON output', {
-                                    rawOutput: stdout,
-                                });
-                                reject(new Error(`Failed to parse Gemini JSON output: ${stdout.slice(0, 200)}`));
-                            }
-                        }
-                        else {
-                            if (stderr.includes('RESOURCE_EXHAUSTED')) {
-                                reject(new Error(`Gemini quota exceeded. Consider using gemini-2.0-flash model. Error: ${stderr.trim()}`));
-                            }
-                            else {
-                                reject(new Error(`Gemini CLI exited with code ${code ?? -1}. Error: ${stderr.trim()}`));
-                            }
-                        }
-                    });
-                    child.on('error', (err) => {
-                        logCliDebug('Failed to spawn gemini CLI', { error: err.message });
-                        reject(new Error(`Failed to spawn gemini CLI. Is it installed and in PATH? Error: ${err.message}`));
-                    });
+                    const parsed = parseGeminiJson(stdout);
+                    if (!parsed.response) {
+                        throw new Error('No response found in Gemini JSON output');
+                    }
+                    return {
+                        response: parsed.response,
+                        usage: null,
+                        threadId: parsed.sessionId,
+                    };
                 }
                 catch (err) {
-                    reject(new Error(`Synchronous error while trying to spawn gemini: ${err instanceof Error ? err.message : String(err)}`));
+                    if (err instanceof Error && err.message.includes('No response')) {
+                        throw err;
+                    }
+                    logCliDebug('Failed to parse Gemini JSON output', {
+                        rawOutput: stdout,
+                    });
+                    throw new Error(`Failed to parse Gemini JSON output: ${stdout.slice(0, 200)}`);
+                }
+            }
+            else {
+                if (stderr.includes('RESOURCE_EXHAUSTED')) {
+                    throw new Error(`Gemini quota exceeded. Consider using gemini-2.0-flash model. Error: ${stderr.trim()}`);
                 }
-            });
+                throw new Error(`Gemini CLI exited with code ${code ?? -1}. Error: ${stderr.trim()}`);
+            }
         },
     };
 }

package/dist/llm-cost.js

@@ -11,6 +11,10 @@ const MODEL_PRICING = {
         inputCostPerMillion: 2.0,
         outputCostPerMillion: 12.0,
     },
+    'gemini-3.1-pro-preview': {
+        inputCostPerMillion: 2.0,
+        outputCostPerMillion: 12.0,
+    },
     'deepseek-reasoner': {
         inputCostPerMillion: 0.55,
         outputCostPerMillion: 2.19,

package/dist/llm.test.js

@@ -18,7 +18,7 @@ vi.mock('./logger.js', () => ({
     logCliDebug: logCliDebugMock,
     logToFile: vi.fn(),
 }));
-vi.mock('child_process', () => ({ spawn: spawnMock }));
+vi.mock('node:child_process', () => ({ spawn: spawnMock }));
 vi.mock('openai', () => {
     class MockOpenAI {
         chat = {
@@ -35,7 +35,9 @@ vi.mock('openai', () => {
 });
 const createChildProcess = () => {
     const child = new EventEmitter();
-    child.stdout = new EventEmitter();
+    const stdout = new EventEmitter();
+    stdout.setEncoding = vi.fn();
+    child.stdout = stdout;
     child.stderr = new EventEmitter();
     child.kill = vi.fn();
     return child;

package/dist/logger.js

@@ -64,7 +64,7 @@ export function logConfiguration(config) {
 }
 export function logCliDebug(message, data) {
     const logMessage = data
-        ? `Gemini CLI: ${message}\n${JSON.stringify(data, null, 2)}`
-        : `Gemini CLI: ${message}`;
+        ? `CLI: ${message}\n${JSON.stringify(data, null, 2)}`
+        : `CLI: ${message}`;
     logToFile(logMessage);
 }

package/dist/models.d.ts

@@ -1 +1 @@
-export declare const ALL_MODELS: readonly ["gemini-2.5-pro", "gemini-3-pro-preview", "deepseek-reasoner", "gpt-5.2", "gpt-5.3-codex", "gpt-5.2-codex"];
+export declare const ALL_MODELS: readonly ["gemini-2.5-pro", "gemini-3-pro-preview", "gemini-3.1-pro-preview", "deepseek-reasoner", "gpt-5.2", "gpt-5.3-codex", "gpt-5.2-codex"];

package/dist/models.js

@@ -1,6 +1,7 @@
 export const ALL_MODELS = [
     'gemini-2.5-pro',
     'gemini-3-pro-preview',
+    'gemini-3.1-pro-preview',
     'deepseek-reasoner',
     'gpt-5.2',
     'gpt-5.3-codex',

package/dist/schema.d.ts

@@ -34,7 +34,7 @@ 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 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.\n\nFor multi-turn conversations with CLI backends (Codex, Gemini CLI, Cursor CLI), the response includes a [thread_id:xxx] prefix. Extract this ID and pass it as the thread_id parameter in follow-up requests to maintain conversation context.";
+    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: Do NOT paste file contents into the prompt field. File contents are automatically read and included by the server when you pass file paths in the `files` parameter. The prompt should only contain your question or instructions.\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.\n\nFor multi-turn conversations with CLI backends (Codex, Gemini CLI, Cursor CLI), the response includes a [thread_id:xxx] prefix. Extract this ID and pass it as the thread_id parameter in follow-up requests to maintain conversation context.";
     readonly inputSchema: z.core.ZodStandardJSONSchemaPayload<z.ZodObject<{
         files: z.ZodOptional<z.ZodArray<z.ZodString>>;
         prompt: z.ZodString;

package/dist/schema.js

@@ -55,6 +55,8 @@ export const toolSchema = {
 
 Be specific about what you want: code implementation, code review, bug analysis, architecture advice, etc.
 
+IMPORTANT: Do NOT paste file contents into the prompt field. File contents are automatically read and included by the server when you pass file paths in the \`files\` parameter. The prompt should only contain your question or instructions.
+
 IMPORTANT: 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.
 
 For multi-turn conversations with CLI backends (Codex, Gemini CLI, Cursor CLI), the response includes a [thread_id:xxx] prefix. Extract this ID and pass it as the thread_id parameter in follow-up requests to maintain conversation context.`,

package/dist/version.d.ts

@@ -1 +1 @@
-export declare const GIT_HASH = "415c21f";
+export declare const GIT_HASH = "1c9e0da";

package/dist/version.js

@@ -1 +1 @@
-export const GIT_HASH = "415c21f";
+export const GIT_HASH = "1c9e0da";

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "consult-llm-mcp",
-  "version": "2.4.0",
+  "version": "2.4.2",
   "description": "MCP server for consulting powerful AI models",
   "type": "module",
   "main": "dist/main.js",