/**
* chat MCP Tool
* Chat with xAI's Grok models
* Based on: https://docs.x.ai/docs/guides/chat-completions
*/
import { z } from "zod";
import { getXAIClient, ChatMessage } from "../xai-client.js";
export const chatSchema = z.object({
message: z
.string()
.describe("The user message to send to Grok"),
model: z
.string()
.optional()
.default("grok-3")
.describe("Chat model: grok-4, grok-3, grok-3-mini, grok-3-fast"),
system_prompt: z
.string()
.optional()
.describe("Optional system/developer prompt to set context (must be first message)"),
temperature: z
.number()
.min(0)
.max(2)
.optional()
.default(0.7)
.describe("Sampling temperature (0-2). Lower = more focused, higher = more creative"),
max_tokens: z
.number()
.optional()
.describe("Maximum tokens in response"),
top_p: z
.number()
.min(0)
.max(1)
.optional()
.describe("Nucleus sampling parameter (0-1)"),
frequency_penalty: z
.number()
.min(-2)
.max(2)
.optional()
.describe("Penalty for token frequency (-2 to 2)"),
presence_penalty: z
.number()
.min(-2)
.max(2)
.optional()
.describe("Penalty for token presence (-2 to 2)"),
});
export type ChatInput = z.infer<typeof chatSchema>;
export const chatTool = {
name: "chat",
description:
"Chat with xAI's Grok models. Send messages and receive AI-generated responses. " +
"Supports summarizing, creative writing, Q&A, coding assistance, and more.",
inputSchema: {
type: "object" as const,
properties: {
message: {
type: "string",
description: "The user message to send to Grok",
},
model: {
type: "string",
description: "Chat model: grok-4, grok-3, grok-3-mini, grok-3-fast",
default: "grok-3",
},
system_prompt: {
type: "string",
description: "Optional system prompt to set context",
},
temperature: {
type: "number",
description: "Sampling temperature (0-2)",
default: 0.7,
},
max_tokens: {
type: "number",
description: "Maximum tokens in response",
},
top_p: {
type: "number",
description: "Nucleus sampling parameter (0-1)",
},
frequency_penalty: {
type: "number",
description: "Penalty for token frequency (-2 to 2)",
},
presence_penalty: {
type: "number",
description: "Penalty for token presence (-2 to 2)",
},
},
required: ["message"],
},
};
export async function handleChat(input: ChatInput): Promise<string> {
const client = getXAIClient();
const validated = chatSchema.parse(input);
const messages: ChatMessage[] = [];
// System/developer message must be first and singular
if (validated.system_prompt) {
messages.push({ role: "system", content: validated.system_prompt });
}
messages.push({ role: "user", content: validated.message });
const response = await client.chatCompletion({
model: validated.model,
messages,
temperature: validated.temperature,
max_tokens: validated.max_tokens,
top_p: validated.top_p,
frequency_penalty: validated.frequency_penalty,
presence_penalty: validated.presence_penalty,
});
const reply = response.choices[0]?.message?.content || "No response";
return JSON.stringify(
{
success: true,
model: response.model,
response: reply,
usage: response.usage,
finish_reason: response.choices[0]?.finish_reason,
},
null,
2
);
}
