/**
* VibeWatch MCP Server
*
* Provides MCP tools for Claude to query terminal output
* Runs on stdio transport
*/
import { Server } from '@modelcontextprotocol/sdk/server/index.js';
import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js';
import {
CallToolRequestSchema,
ListToolsRequestSchema,
} from '@modelcontextprotocol/sdk/types.js';
const API_BASE = 'http://localhost:3333';
// Create MCP server
const server = new Server(
{
name: 'vibewatch',
version: '0.1.0',
},
{
capabilities: {
tools: {},
},
}
);
// List available tools
server.setRequestHandler(ListToolsRequestSchema, async () => {
return {
tools: [
{
name: 'ping',
description: 'Test tool - responds with pong',
inputSchema: {
type: 'object',
properties: {},
},
},
{
name: 'get_terminal_output',
description: 'Get the last N lines of terminal output from a monitored development process. Use this to see what\'s happening in the terminal, check for errors, or understand the current state of a running process.',
inputSchema: {
type: 'object',
properties: {
lines: {
type: 'number',
description: 'Number of lines to retrieve (max 100)',
default: 50,
},
filter: {
type: 'string',
enum: ['all', 'errors', 'warnings'],
description: 'Filter output to specific types. Use "errors" to see only error messages, "warnings" for warnings, or "all" for everything.',
default: 'all',
},
detail: {
type: 'string',
enum: ['errors', 'context', 'full'],
description: 'Detail level for progressive disclosure. "errors" returns only error lines (~200 tokens), "context" returns errors with surrounding context (~500 tokens), "full" returns all output (~1000 tokens). Start with "errors" or "context" to save tokens.',
default: 'full',
},
},
},
},
{
name: 'get_crash_context',
description: 'Get detailed context about a process crash, including the error message, stack trace, and relevant file paths. Use this when the user mentions a crash or when you need to understand why a process failed.',
inputSchema: {
type: 'object',
properties: {
verbose: {
type: 'boolean',
description: 'Include node_modules/site-packages in stack traces (default: false)',
default: false,
},
},
},
},
{
name: 'get_recent_errors',
description: 'Get only recent error lines from the terminal output. This is a quick way to see what went wrong without scrolling through all output.',
inputSchema: {
type: 'object',
properties: {},
},
},
],
};
});
// Handle tool calls
server.setRequestHandler(CallToolRequestSchema, async (request) => {
const { name, arguments: args } = request.params;
if (name === 'ping') {
return {
content: [
{
type: 'text',
text: 'pong',
},
],
};
}
if (name === 'get_terminal_output') {
const lines = Math.min((args?.lines as number) || 50, 100);
const filter = (args?.filter as string) || 'all';
const detail = (args?.detail as string) || 'full';
try {
const response = await fetch(
`${API_BASE}/live?lines=${lines}&filter=${filter}&detail=${detail}`
);
if (!response.ok) {
return {
content: [
{
type: 'text',
text: JSON.stringify({
error: 'No monitored process running',
hint: 'Start a process with: vibewatch <command>',
}),
},
],
};
}
const data = await response.json();
return {
content: [
{
type: 'text',
text: JSON.stringify(data, null, 2),
},
],
};
} catch (error) {
return {
content: [
{
type: 'text',
text: JSON.stringify({
error: 'Failed to connect to VibeWatch',
hint: 'Make sure you have a process running with: vibewatch <command>',
details: error instanceof Error ? error.message : String(error),
}),
},
],
};
}
}
if (name === 'get_crash_context') {
const verbose = (args?.verbose as boolean) ?? false;
try {
const response = await fetch(
`${API_BASE}/crash?verbose=${verbose}`
);
if (!response.ok) {
return {
content: [
{
type: 'text',
text: JSON.stringify({
error: 'No monitored process running',
hint: 'Start a process with: vibewatch <command>',
}),
},
],
};
}
const data = await response.json() as { error?: string; [key: string]: unknown };
// Check if there's no crash
if (data.error === 'No crash detected') {
return {
content: [
{
type: 'text',
text: JSON.stringify({
status: 'no_crash',
message: 'The process is still running or exited cleanly.',
hint: 'Use get_terminal_output to see the current output.',
}),
},
],
};
}
return {
content: [
{
type: 'text',
text: JSON.stringify(data, null, 2),
},
],
};
} catch (error) {
return {
content: [
{
type: 'text',
text: JSON.stringify({
error: 'Failed to connect to VibeWatch',
hint: 'Make sure you have a process running with: vibewatch <command>',
details: error instanceof Error ? error.message : String(error),
}),
},
],
};
}
}
if (name === 'get_recent_errors') {
try {
const response = await fetch(`${API_BASE}/live?filter=errors`);
if (!response.ok) {
return {
content: [
{
type: 'text',
text: JSON.stringify({
error: 'No monitored process running',
hint: 'Start a process with: vibewatch <command>',
}),
},
],
};
}
const data = await response.json() as { output?: string[]; process_status?: string; [key: string]: unknown };
if (data.output?.length === 0) {
return {
content: [
{
type: 'text',
text: JSON.stringify({
status: 'no_errors',
message: 'No errors detected in recent output.',
process_status: data.process_status,
}),
},
],
};
}
return {
content: [
{
type: 'text',
text: JSON.stringify(data, null, 2),
},
],
};
} catch (error) {
return {
content: [
{
type: 'text',
text: JSON.stringify({
error: 'Failed to connect to VibeWatch',
hint: 'Make sure you have a process running with: vibewatch <command>',
details: error instanceof Error ? error.message : String(error),
}),
},
],
};
}
}
throw new Error(`Unknown tool: ${name}`);
});
// Error handling
server.onerror = (error) => {
console.error('[VIBE-WATCH MCP] Error:', error);
};
// Start server
async function main() {
const transport = new StdioServerTransport();
await server.connect(transport);
console.error('[VIBE-WATCH MCP] Server ready on stdio transport');
}
main().catch((error) => {
console.error('[VIBE-WATCH MCP] Fatal error:', error);
process.exit(1);
});
