@miraigent/free-ai-ops-mcp 0.1.34 → 0.1.36

package/CHANGELOG.md

@@ -1,5 +1,21 @@
 # Changelog
 
+## [0.1.36] - 2026-07-11
+
+### Added
+
+- Added smoke-test coverage that verifies unknown MCP tool calls return a
+  JSON-RPC error with the request ID instead of being treated as a successful
+  tool result.
+
+## [0.1.35] - 2026-07-10
+
+### Changed
+
+- Added a README version mismatch checklist so npm and GitHub users can compare
+  the registry version, npx help output, and MCP initialize response before
+  changing desktop client configuration.
+
 ## [0.1.34] - 2026-07-09
 
 ### Changed

package/README.md

@@ -63,6 +63,18 @@ public registry version before changing private client configuration:
 The version printed by `npm view` should match the version shown in the package
 help output and in the MCP `initialize` response.
 
+Quick version mismatch check:
+
+1. Run `npm view @miraigent/free-ai-ops-mcp version`.
+2. Run `npx -y free-ai-ops-mcp@npm:@miraigent/free-ai-ops-mcp --help` and
+   confirm the header shows the same version.
+3. Run the one-command smoke test and confirm the `initialize` response includes
+   the same `serverInfo.version`.
+
+If only the desktop MCP client shows an older version after those checks pass,
+restart the client and re-check its `command` and `args` path before changing
+any private workflow data.
+
 If the result is unclear, open a Tried It feedback issue with synthetic input
 only. A useful issue can be as small as:
 

package/examples/mcp-json-rpc-session/README.md

@@ -44,6 +44,15 @@ In raw JSON-RPC output, the tool result is wrapped by MCP as `content[0].text`.
 Parse that text value as JSON, then check `gateStatus` before connecting a real
 workflow or changing a desktop client config.
 
+For a terminal-only check, you can read just the final response line from the
+public sample and print the parsed tool result:
+
+    npm run mcp < examples/mcp-json-rpc-session/sample-session.jsonl | tail -n 1 | node -e 'process.stdin.on("data", c => { const r = JSON.parse(c); console.log(JSON.parse(r.result.content[0].text)); })'
+
+That should print a public synthetic `human_review_gate` result with
+`gateStatus: "stop"`. Use only the bundled sample input for this quick check,
+not private customer records or desktop client logs.
+
 If those fields appear in the terminal but not in Claude Desktop, Cursor, or
 another client, the package is likely running and the next thing to check is the
 client config, restart behavior, or command path.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@miraigent/free-ai-ops-mcp",
-  "version": "0.1.34",
+  "version": "0.1.36",
   "description": "Free MCP server for developers adding human review gates, prompt risk checks, FAQ review, and CRM note safety to AI tools.",
   "homepage": "https://github.com/Miraigent/miraigent-free-ai-ops-mcp",
   "type": "module",

package/scripts/smoke-test.mjs

@@ -205,3 +205,16 @@ assert.equal(
   promptRiskToolResult.boundary,
   'This tool is a prompt risk helper. It is not legal advice and does not call an AI API.'
 );
+
+const unknownToolResponses = await runServerWithRequests([
+  JSON.stringify({
+    jsonrpc: '2.0',
+    id: 7,
+    method: 'tools/call',
+    params: { name: 'unknown_public_tool', arguments: {} }
+  })
+]);
+assert.equal(unknownToolResponses.length, 1);
+assert.equal(unknownToolResponses[0].id, 7);
+assert.equal(unknownToolResponses[0].error.code, -32000);
+assert.equal(unknownToolResponses[0].error.message, 'Unknown tool: unknown_public_tool');