// MCP TOOL REFERENCE

THOUGHTBOX TOOLS

85 tools from the Thoughtbox MCP Server, categorised by risk level.

// ALL 85 THOUGHTBOX TOOLS

READ 68 tools

Read analysis Deep analysis of reasoning sessions Read ARCHITECT Structural design specialist Read architecture System design, component relationships, structure Read branch Toolhost for managing reasoning branches. Read branch-management Spawn and merge reasoning branches Read branch-retrieval List and inspect branches Read budget Energy budget for implementation (default: 100) Read communication Explaining clearly to humans, documentation Read DEBUGGER Root cause analysis specialist Read debugging Finding and fixing issues in code, logic, or systems Read decision The decision to make, or the options to evaluate Read decision-making Choosing between options under uncertainty Read deep Perform deeper semantic analysis and cross-file trace (default: false) Read depth Spec depth: shallow, standard, or comprehensive (default: standard) Read estimation Making reasonable guesses with limited information Read evolution-check Get instructions for checking which prior thoughts should be updated when a new insight is added. Uses sub-... Read focus Optional area to focus on (e.g., Read hub_reply Post a message to a problem Read initialization Stage progression and cipher loading Read input Your task/question. May include Read insight The new insight that may require updating prior thoughts Read interleaved-thinking Use this Thoughtbox server as a reasoning workspace to alternate between internal reasoning steps and exter... Read list_mcp_assets Overview of all MCP capabilities, tools, resources, and quickstart guide Read MANAGER Delegation and team coordination specialist Read max_iterations Maximum iterations per spec (default: 3) Read max_specs Maximum number of specs to generate (default: 5) Read navigation Browse and navigate the project/task/aspect hierarchy Read output_folder Folder to output specs to (default: .specs/) Read planning Breaking down work, sequencing tasks, project organization Read prioritization Deciding what to do first, resource allocation Read problem Description of the unexpected behavior or bug Read prompt The feature or system to design specifications for Read prompt_or_spec_path Prompt string OR existing spec folder/file path Read reasoning Record, read, and explore thought chains Read report_only Skip suggestion phase, only generate report (default: false) Read RESEARCHER Parallel hypothesis investigation specialist Read REVIEWER Code and proposal review specialist Read risk-analysis Identifying what could go wrong and how to prevent it Read scope Files to refactor and description of structural change Read SECURITY Risk and vulnerability detection specialist Read session Toolhost for managing Thoughtbox reasoning sessions. List, search, retrieve, resume, export, analyze, and e... Read session-analysis Analyze session structure and extract learnings for DGM evolution Read session-management Resume and export sessions Read session-retrieval List, search, and retrieve session details Read sessionId Session ID to review, or Read skip_design Treat input as existing specs, skip design phase (default: false) Read skip_validation Skip validation, jump from design to orchestration (default: false) Read spec_folder Folder containing specification documents to implement Read spec_path Path to spec file or folder to validate Read spec-designer Design and produce implementation specifications through structured cognitive loops. Creates specs from pro... Read spec-orchestrator Coordinate implementation of multiple specification documents from a folder. Manages dependencies, tracks p... Read spec-validator Systematically validate specification documents against current codebase and project architecture. Identifi... Read specification-suite Chain the design, validate, orchestrate lifecycle into one command. Moves from blank prompt to implemented,... Read strict Strict mode: any missing requirement is a failure (default: false) Read subagent-summarize Get instructions for using Claude Code Read task The task to complete using interleaved thinking approach Read test-memory Behavioral tests for the thoughtbox_knowledge tool (12 tests covering entities, observations, relations, gr... Read test-mental-models Behavioral tests for the mental_models structured reasoning tool (6 tests covering discovery, retrieval, ca... Read test-notebook Behavioral tests for the notebook literate programming tool (8 tests covering creation, cells, execution, e... Read test-thoughtbox Behavioral tests for the thoughtbox thinking tool (15 tests covering forward/backward thinking, branching, ... Read thoughtbox_branch Toolhost for managing reasoning branches. Read thoughtbox_gateway Always-available routing tool for Thoughtbox operations. All operations use the same flat shape: { operation: Read thoughtbox_knowledge Knowledge graph operations for tracking entities, relationships, and temporal observations. Read thoughtbox_search Discover Thoughtbox operations, prompts, and resources by writing JavaScript that queries the catalog. The \ Read thoughtbox_session Toolhost for managing Thoughtbox reasoning sessions. List, search, retrieve, resume, export, analyze, and e... Read thoughts_limit Maximum number of thoughts to use (default: 100) Read topic The research question or topic to investigate Read validation Checking assumptions, testing hypotheses, verification

WRITE 8 tools

Write cell-operations Add, update, list, and retrieve cells Write entity-management Create, read, and observe entities in the knowledge graph Write graph-structure Create relations and traverse the graph Write init Navigate and manage Thoughtbox sessions. Operations: - get_state: Get current navigation state and availab... Write notebook-management Create, list, load, and export notebooks Write plan_only Only explore and plan, don Write session-setup Load existing sessions or start new work Write thoughtbox_init Initialize and navigate Thoughtbox sessions, load context, or manage project bounds.

DESTRUCTIVE 2 tools

Destructive clear_folder Whether to clear the .interleaved-thinking folder after completion, keeping only final-answer.md (default: ... Destructive thoughtbox_ulysses Ulysses Protocol: surprise-gated debugging for autonomous agents. Forces pre-committed recovery, rigorous s...

EXECUTE 7 tools

Execute execution Run code cells and install dependencies Execute hub_action Execute a quick Hub action: claim a problem, update status, endorse consensus, or review a proposal Execute notebook Notebook toolhost for literate programming with JavaScript/TypeScript. Create, manage, and execute interac... Execute thoughtbox_execute Run JavaScript using the \ Execute thoughtbox_notebook Notebook toolhost for literate programming with JavaScript/TypeScript. Create, manage, and execute interact... Execute thoughtbox_theseus Theseus Protocol: friction-gated refactoring for autonomous agents. Prevents scope drift via boundary locki... Execute thoughtbox_thought Advanced reasoning tracking tool. Submit thoughts, track state changes, audit decisions, and build branches...

// RUNNING THIS SERVER

The managed route: connect Thoughtbox through the PolicyLayer gateway — every tool call above is checked against your policy before it runs, with a full audit log.

DIRECT INSTALL (UNMANAGED) npx -y @kastalien-research/thoughtbox

// FAQ

How many tools does the Thoughtbox MCP server have? +

The Thoughtbox MCP server exposes 85 tools across 4 categories: Read, Write, Destructive, Execute.

How do I enforce policies on Thoughtbox tools? +

Route the Thoughtbox server through the PolicyLayer gateway. Define allow, deny, or approval rules per tool in the dashboard — they are enforced on every call before it reaches the server.

What risk categories do Thoughtbox tools fall into? +

Thoughtbox tools are categorised as Read (68), Write (8), Destructive (2), Execute (7). Each category has a recommended default policy.

Let agents act without letting them run wild.

Route your MCP servers through PolicyLayer and every tool call is checked against your policy before it runs — allow, deny, or require approval. Per-identity grants. Full audit log. Live in minutes.

SECURE YOUR MCP →

Free to start. No card required.

4,600+ MCP servers and 31,000+ tools scanned and risk-classified.