Critical Thinking

A Model Context Protocol server for critical, narrated, sequential thinking. Think one step at a time, out loud — with required confidence calibration and adversarial self-critique on every thought.

It fuses three disciplines:

1. Sequential thinking — break problems into ordered, numbered steps; revise; branch.
2. Thinking out loud — explain each thought in first-person, exploratory voice. Putting half-formed reasoning into words is itself the double-check on it.
3. Critical self-examination — every thought is paired with confidence, assumptions, critique, and a counter-argument.

The single tool is criticalthinking. Every call must include the four critical-thinking fields — there is no opt-out, by design.

Install

go install github.com/jacaudi/critical-thinking/cmd/critical-thinking@latest
# or
docker pull ghcr.io/jacaudi/critical-thinking:v1.4.2

The Go install lands the binary at $GOPATH/bin/critical-thinking.

Run

# stdio (default; for Claude Desktop, Codex CLI, VS Code, etc.)
critical-thinking

# Streamable HTTP
critical-thinking -http :3000

# Docker (HTTP on :3000)
docker run --rm -p 3000:3000 ghcr.io/jacaudi/critical-thinking:v1.4.2

One-call example

Request:

{
  "thought": "I think we should normalize first because reads dominate writes.",
  "thoughtNumber": 1, "totalThoughts": 3, "nextThoughtNeeded": true,
  "confidence": 0.6,
  "assumptions": ["read:write ratio is ~10:1"],
  "critique": "Drifted into solution mode without confirming the ratio.",
  "counterArgument": "If writes dominate, monolith-first is simpler.",
  "nextStepRationale": "Verify the read:write ratio before committing to normalization."
}

Response (structuredContent):

{ "branches": [], "thoughtHistoryLength": 1, "sessionConfidence": 0.6 }

The text content is a rendered transcript in first-person, exploratory voice. Subsequent calls can omit thoughtNumber (auto-assigned) and totalThoughts (inherited). Every critical-thinking field has a server-side length cap to enforce one-tight-sentence discipline. The full contract lives in the tool description itself.

Client setup

Claude Code

Add the server with the claude CLI — pick one transport.

stdio (the binary runs as a subprocess of Claude Code):

claude mcp add critical-thinking -- critical-thinking

Streamable HTTP (run the server separately, point Claude Code at the URL):

critical-thinking -http :3000 &
claude mcp add --transport http critical-thinking http://localhost:3000/mcp

Scope it with --scope user (available in every project for your user) or --scope project (committed to .mcp.json in the repo). Default scope is local (this project, your machine only).

Verify with claude mcp list. Inside a session, /mcp shows server status and tools.

Other clients

mcp.json (Claude Desktop / Codex CLI / VS Code):

{
  "mcpServers": {
    "critical-thinking": { "command": "critical-thinking" }
  }
}

Or HTTP:

{
  "mcpServers": {
    "critical-thinking": { "url": "http://localhost:3000/mcp" }
  }
}

More client recipes in docs/clients.md.

Resources

The server exposes thinking://current — a per-session JSON snapshot of the full thought history (trunk + branches, all critical-thinking fields preserved).

Documentation

• docs/configuration.md — env vars, HTTP endpoints, session lifecycle
• docs/clients.md — Claude Desktop, Codex CLI, VS Code, Cursor recipes
• docs/development.md — building, testing, debugging with MCP Inspector
• docs/migration.md — breaking changes since http-sequential-thinking

License

MIT.