threadctx-mcp 0.1.1 → 0.1.2

package/dist/server.js

@@ -1,3 +1,6 @@
+import { readFileSync } from 'node:fs';
+import { dirname, join } from 'node:path';
+import { fileURLToPath } from 'node:url';
 import { Server } from '@modelcontextprotocol/sdk/server/index.js';
 import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js';
 import { CallToolRequestSchema, ListToolsRequestSchema } from '@modelcontextprotocol/sdk/types.js';
@@ -8,7 +11,30 @@ import { CloudClient } from './cloud-client.js';
 // output, Cursor's agent panel, future surfaces). See spec section 2.4 —
 // the same short string everywhere is what makes the brand legible.
 const attributionFooter = (n) => `· via threadctx — shared team memory (${n} hit${n === 1 ? '' : 's'})`;
+// package.json sits next to dist/ both in this repo and once installed as a
+// dependency, so this resolves correctly in both dev and published contexts.
+const packageJsonPath = join(dirname(fileURLToPath(import.meta.url)), '..', 'package.json');
+const packageVersion = JSON.parse(readFileSync(packageJsonPath, 'utf-8')).version;
 export async function startServer() {
+    console.error('[threadctx] Shared memory MCP server for Claude Code, Cursor, and other MCP clients.');
+    console.error('[threadctx] Setup guide: https://threadctx.dev');
+    // A human typing `npx threadctx-mcp` straight into a terminal (rather than
+    // an MCP client spawning it over a pipe) is a very natural thing to try
+    // when kicking the tires — and without this check it just sits there
+    // forever waiting for JSON-RPC input that will never arrive, with zero
+    // explanation. Detect that case and exit with guidance instead of hanging.
+    if (process.stdin.isTTY) {
+        console.error('');
+        console.error("[threadctx] This looks like a terminal, not an MCP client — there's nothing more to do here.");
+        console.error('[threadctx] threadctx only runs as a subprocess that Claude Code / Cursor launch for you.');
+        console.error('[threadctx] Add this to your MCP client config (~/.claude/mcp.json or .cursor/mcp.json):');
+        console.error('');
+        console.error(JSON.stringify({ mcpServers: { threadctx: { command: 'npx', args: ['-y', 'threadctx-mcp'] } } }, null, 2));
+        console.error('');
+        console.error('[threadctx] Then restart your agent — it will pick up memory_write / memory_query.');
+        console.error('[threadctx] Full guide: https://threadctx.dev');
+        process.exit(0);
+    }
     const config = loadConfig();
     if (config.mode === 'cloud' && !config.apiKey) {
         console.error('[threadctx] THREADCTX_MODE=cloud but no API key was found. Falling back to local mode. ' +
@@ -18,7 +44,7 @@ export async function startServer() {
     const localStore = useCloud ? null : new LocalStore(config.dbPath);
     const cloudClient = useCloud ? new CloudClient(config.apiUrl, config.apiKey, config.actorId) : null;
     const repo = config.repo;
-    const server = new Server({ name: 'threadctx', version: '0.1.0' }, { capabilities: { tools: {} } });
+    const server = new Server({ name: 'threadctx', version: packageVersion }, { capabilities: { tools: {} } });
     server.setRequestHandler(ListToolsRequestSchema, async () => ({
         tools: [
             {
@@ -106,5 +132,8 @@ export async function startServer() {
     });
     const transport = new StdioServerTransport();
     await server.connect(transport);
-    console.error(`[threadctx] MCP server running in ${useCloud ? 'cloud' : 'local'} mode for repo "${repo}".`);
+    const repoSuffix = repo === 'unknown-repo'
+        ? ' (no git remote found in this directory — cd into a repo with one, or set THREADCTX_REPO)'
+        : '';
+    console.error(`[threadctx] MCP server running in ${useCloud ? 'cloud' : 'local'} mode for repo "${repo}"${repoSuffix}.`);
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "threadctx-mcp",
-  "version": "0.1.1",
+  "version": "0.1.2",
   "description": "Shared memory MCP server for AI coding agents. Local-only by default; point it at threadctx.dev (or your own deployment) to share memory across your team.",
   "type": "module",
   "bin": {