Third-party toolUpdated 2026.11.14

Sequential Thinking MCP: Test Steps, Revisions, and Validation

Test the reference Sequential Thinking MCP 2.1.1 with two steps, a revision, and an invalid counter. A successful tool call does not prove reasoning quality.

Aaron recommends Open official GitHub

Sequential Thinking MCP provides a structured tool for recording steps, revisions, and branches. This test of 2.1.1 accepted two steps and a revision, then rejected thoughtNumber=0. No model was involved. Successful responses therefore establish tool behavior, not better answers or verified reasoning quality.

To save results as queryable entity relationships, compare Memory MCP's persistent data. If a call returns unexpected output, inspect parameters and responses with MCP Inspector.

Pin the server and connection scope

This page covers the reference implementation in modelcontextprotocol/servers, published as @modelcontextprotocol/server-sequential-thinking. Version 2.1.1 requires Node.js 24 or later. The fixture client used a 2025-03-26 initialization over stdio, with no Claude, Cursor, or other AI-host login.

Send explicit step data

The first record used thoughtNumber=1, totalThoughts=2, and nextThoughtNeeded=true with a synthetic dispatch checklist statement. The second used thoughtNumber=2 and nextThoughtNeeded=false. A third call added isRevision=true and revisesThought=1 to exercise revision fields.

These commands use a POSIX shell. The test ran in Linux; Windows shell behavior was not tested.

npm install --ignore-scripts --save-exact @modelcontextprotocol/server-sequential-thinking@2.1.1
# POSIX shell
DISABLE_THOUGHT_LOGGING=true node node_modules/@modelcontextprotocol/server-sequential-thinking/dist/index.js

Exercise valid and invalid input

All three valid calls completed without a JSON-RPC error or tool isError. Changing thoughtNumber to zero produced a tool error. Integration should inspect that error rather than equating a running stdio process with successful calls. Counters in the response describe tool state; they are not external evidence for an answer.

Decide whether it helps your workflow

To evaluate model adoption, use a task with a verifiable deliverable and separately check actual tool calls and final results. Branching, long-session state, newer-protocol behavior, and host configuration are documented from official v2.1.1 material and were not locally tested. The lab set DISABLE_THOUGHT_LOGGING=true and used synthetic checklist statements rather than private work conversations.

Official sources and test scope

Codex ran this synthetic-data test in a credential-free, non-root, network-disabled container. It is not a claim that the site owner tested it. Package version: 2.1.1; documentation snapshot: e1dc4b18dbb9.

Subscribe to AI tool picks and practical advice

Get worthwhile tool picks, test findings and important updates in your inbox.

Get a monthly-updated list of AI tools and Skills right away. We email when there are worthwhile tool picks, test findings or important updates. Unsubscribe anytime. Privacy

Almost done: check your inboxClick the button in the confirmation email and you will get the download link for the tool list. If you do not see it, check your spam or Promotions folder.

Or follow us on: