The most advanced sequential reasoning MCP for autonomous AI workflows

Enables structured problem-solving with parallel exploration, branch navigation, and persistent thought chains that survive context window compaction.

Built by a developer working with AI to solve real autonomous reasoning challenges. Extends dormant branching capabilities in Anthropic's original sequential-thinking server.

Why we created MAXential Thinking

Reasoning is Ephemeral: AI model, LLM, reasoning is ephemeral. When context windows fill up, thinking gets compressed or is lost. Complex problems and brainstorming need exploration of multiple ideas and approaches. We wanted a way to optimize AI's abilities by having them think in parallel, with multisteps, backtracking when needed, and to summarize their findings.

Branching Chains: While designing MAXential Thinking to enable reasoning through multiple thought-chains, we looked to Anthropic's sequential-thinking server for inspiration. We discovered it had branching capabilities in its schema but no tools to access them. So we forked it and expanded their dormant branching with 11 tools that allow AI to actively reason.

For AI Assistants (Claude, ChatGPT, etc.)

Using this MCP server instead of Sequential Thinking MCP, you gain:

Structured Reasoning That Persists

• Thought chains survive context window compaction
• Navigate back to earlier reasoning without losing context
• Revise earlier thoughts as understanding evolves
• Export complete reasoning chains for handoff or documentation

Parallel Exploration

• Branch to explore different approaches/ideas/thoughts simultaneously
• Compare solutions/outcomes side-by-side before choosing
• Merge insights back to main reasoning thread
• Close branches that come up empty without impacting main thread

Autonomous Workflow Enhancement

• Tag thoughts for semantic organization (hypothesis, evidence, decision, finding)
• Search your reasoning history by content or tags
• Generate diagrams of your thought process (Mermaid/ASCII)
• Build complex solutions with transparent reasoning

What Makes This Different

Original sequential-thinking: 1 tool with 9 parameters. Branch parameters existed but no way to use them.

MAXential Thinking v2.3: 20 focused tools that match reasoning patterns:

• No parameter juggling - simply call the right tool
• Branching works - explore, switch, merge, close

AI models benefit from structured thinking. This server externalizes that process, making it:

• Transparent - See the reasoning chain, not just conclusions
• Navigable - Branch to explore alternatives, backtrack, revise
• Searchable - Find thoughts by content or tags
• Exportable - Share reasoning as markdown or JSON

Features

Core Thinking (v2.0)

| Tool | Description | |------|-------------| | think | Add a thought to your reasoning chain | | revise | Revise a previous thought | | complete | Mark thinking chain complete with conclusion |

Branching (v2.0)

| Tool | Description | |------|-------------| | branch | Create a new reasoning branch (like git) | | switch_branch | Switch between branches (use "main" to return) | | list_branches | List all branches with status | | get_branch | Get details of a specific branch | | close_branch | Close a branch with optional conclusion | | merge_branch | Merge insights back to main |

Navigation (v2.0)

| Tool | Description | |------|-------------| | get_thought | Retrieve a specific thought by number | | get_history | Get thought history, optionally filtered |

Organization (v2.2)

| Tool | Description | |------|-------------| | reset | Clear session, start fresh | | tag | Add semantic tags to thoughts (hypothesis, evidence, decision, etc.) | | search | Find thoughts by content or tags | | export | Export to markdown or JSON | | visualize | Generate ASCII or Mermaid diagrams |

Session Persistence (v2.3)

| Tool | Description | |------|-------------| | session_save | Name and describe the current session for later retrieval | | session_load | Restore a saved session — all thoughts, branches, and tags | | session_list | Browse available sessions, filter by status | | session_summary | Generate a compressed summary for token-efficient context loading |

Thinking sessions are automatically persisted to SQLite as you work. Every think, branch, tag, and revise call writes through to disk in real time. Sessions survive server restarts and context window resets — pick up where you left off in a new conversation.

Installation

Claude Desktop

Add to your claude_desktop_config.json:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "maxential-thinking": {
      "command": "npx",
      "args": ["-y", "@bam-devcrew/maxential-thinking-mcp"]
    }
  }
}

From Source

git clone https://github.com/BAM-DevCrew/MAXential-Thinking-MCP.git
cd MAXential-Thinking-MCP
npm install
npm run build

Then configure:

{
  "mcpServers": {
    "maxential-thinking": {
      "command": "node",
      "args": ["/path/to/MAXential-Thinking-MCP/dist/src/index.js"]
    }
  }
}

Configuration

Persistence (v2.3)

Session data is stored in SQLite. By default, the database file is created at .maxential/thinking.db in the working directory. You can customize this:

{
  "mcpServers": {
    "maxential-thinking": {
      "command": "npx",
      "args": ["-y", "@bam-devcrew/maxential-thinking-mcp"],
      "env": {
        "MAXENTIAL_DB_PATH": "/path/to/your/thinking.db"
      }
    }
  }
}

| MAXENTIAL_DB_PATH value | Behavior | |---------------------------|----------| | (not set) | .maxential/thinking.db in working directory (default) | | /path/to/file.db | Use explicit file path | | :memory: | In-memory SQLite — no file created, data lost on restart |

If SQLite initialization fails (permissions, native module issues), the server falls back to in-memory mode automatically — it never crashes.

Note: Add .maxential/ to your project's .gitignore to keep session data out of version control.

Usage Examples

Basic Thinking

think: "The user wants to optimize database queries"
think: "First, I should check for N+1 query problems"
think: "Found 3 N+1 issues in the user service"
tag: thoughtNumber=3, add=["finding", "database"]
complete: "Fixed N+1 queries, reduced load time by 40%"

Branching for Alternatives

think: "Should we use REST or GraphQL?"
branch: branchId="rest-approach", reason="Exploring REST"
think: "REST is simpler, better caching"
switch_branch: branchId="main"
branch: branchId="graphql-approach", reason="Exploring GraphQL"
think: "GraphQL reduces overfetching"
merge_branch: branchId="rest-approach", strategy="summary"

Search and Export

search: query="database"           # Find by content
search: tags=["decision"]          # Find by tag
export: format="markdown"          # Get full chain as markdown
visualize: format="mermaid"        # Get diagram for docs

Session Persistence

think: "Working through the auth redesign..."
think: "JWT approach has these tradeoffs..."
session_save: name="Auth redesign analysis"

# Later, in a new conversation:
session_list:                      # Browse saved sessions
session_load: id="<session-uuid>"  # Restore full context
think: "Continuing where I left off..."
session_summary: id="<session-uuid>"  # Get compressed recap

Visualization Output

Mermaid (for GitHub, Obsidian, etc.)

graph TD
    T1["#1"] --> T2["#2"]
    T1 --> T3["#3"]
    T3 -.revises.-> T1

    subgraph alt-approach
        T2
    end

ASCII

Thinking Chain
==============

T1 --> T2 --> T3
Branch [alt-approach] from T1

Revisions:
  T3 revises T1

Logging

Enable file logging by setting environment variable:

{
  "mcpServers": {
    "maxential-thinking": {
      "command": "node",
      "args": ["/path/to/dist/src/index.js"],
      "env": {
        "MAXENTIAL_LOG_FILE": "/path/to/error.log"
      }
    }
  }
}

Development

npm install          # Install dependencies
npm run build        # Build TypeScript
npm run watch        # Watch mode for development

License

MIT

Contributing

Issues and PRs welcome at github.com/BAM-DevCrew/MAXential-Thinking-MCP