Loading ToolBench
Local-first MCP workspace extension for Claude Desktop with a low-poly isometric mission control facility. Provides project management, task tracking, decision recording, artifact registration, and checkpoint saving with activity event observability.
30 tools with consistent verb-noun naming and comprehensive descriptions (avg 180 chars). All tools have input schemas with typed parameters and descriptions. Strong side-effect documentation (WRITE/READ_ONLY/DESTRUCTIVE risk labels). However, output schemas are not documented in the source code, LLMs cannot see what fields to expect from responses. Error handling guidance is absent; tools do not indicate recovery paths or retryability. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) visible in schema. Composition is sound: tools chain well (e.g., create_project → set_active_project → create_task). Parameter constraints are well-specified (enums for stage/status, length limits in descriptions). Missing: documented return types, pagination guidance for list tools, and actionable error messages.
Applies a bulk task status update using a confirmation token from preview_bulk_task_update. Side effect: persists all task changes and resolves the approval wait.
Applies a project import using a confirmation token from preview_project_import. Side effect: merges imported data into the project and resolves the approval wait.
Approves a pending operation (bulk task update or project import) that is awaiting dashboard approval. Side effect: executes the approved operation and resolves the approval wait.
Archives a project, removing it from the active workspace. Side effect: marks the project as archived in the database.
Creates a new project with an initial name and optional description. Side effect: persists the project to the database and sets it as the active project.
Creates a new task in the active project with an initial title and optional description. Side effect: persists the task to the database.
Output schemas not documented. LLMs cannot see what fields list_projects, get_project_brief, or prepare_project_context return. This forces agents to guess field names and risks chaining failures.
No error handling guidance. Tools do not indicate which errors are retryable, which require user intervention, or what recovery steps to take. Agents cannot self-correct on failure.
List tools lack pagination parameters. list_projects, list_tasks, list_decisions, list_artifacts do not accept limit/offset or return total counts. Large workspaces will blow context windows.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | 2025-06-18+ | v2 |
Packages a project export (all data, tasks, decisions, artifacts, checkpoints) into a portable format. Side effect: writes the export file to the exports directory.
Loads the most recent checkpoint for a project. Read-only; no side effects.
Retrieves the current dashboard state including projects, tasks, activity timeline, and current activity. Read-only; no side effects.
Loads the brief for a project: its name, description, stage, status, task summary, and recent decisions. Read-only; no side effects.
Lists all artifacts registered for a project. Read-only; no side effects.
Lists all decisions recorded for a project. Read-only; no side effects.
Lists all projects in the workspace with their current status, stage, and metadata. Read-only; no side effects.
Lists all tasks in a project with their current status, priority, and metadata. Read-only; no side effects.
Records verification of an artifact (e.g., code review passed, design approved). Side effect: updates the artifact's verification status in the database.
Phase 0 connectivity check for the Claude Mission Control proof of concept. Returns server time and version. Has no side effects.
Opens the Claude Mission Control Phase 0 proof-of-concept dashboard. Shows only saved PoC state and observable PoC tool activity. Read-only.
Prepares a context package containing the project brief, all tasks, decisions, and latest checkpoint for use in Claude's reasoning. Read-only; no side effects.
Previews a bulk task status update before applying it. Returns a confirmation token required to apply the changes. Side effect: opens a Security Gate waiting event for approval.
Previews a project import from an export file before applying it. Returns a confirmation token required to apply the import. Side effect: opens a Security Gate waiting event for approval.
Records a project decision with a title, rationale, and optional alternatives considered. Side effect: persists the decision to the database.
Records a timestamped test event in the local Mission Control PoC state file. Side effect: appends one event (capped history) to local storage. No other data is touched.
Records the result of a validation check (e.g., test pass/fail, compliance check). Side effect: persists the validation result to the database.
Registers an artifact (document, code, design, etc.) associated with a project. Side effect: persists the artifact record to the database.
Rejects a pending operation (bulk task update or project import) that is awaiting dashboard approval. Side effect: cancels the operation and resolves the approval wait.
Saves a project checkpoint with a summary of progress and optional recommended next action. Side effect: persists the checkpoint to the database.
Sets the active project for the workspace. Side effect: updates the active project setting.
Updates a project's name and/or description. Side effect: persists changes to the database.
Advances a project to the next stage in its lifecycle (Scoping → Planning → Executing → Reviewing → Closed). Side effect: persists the stage change and records the transition.
Updates a task's title, description, status, or blocked reason. Side effect: persists changes to the database.
No tool annotations visible. Schemas do not include readOnlyHint, destructiveHint, or idempotentHint. Clients cannot infer which tools are safe to call speculatively or which require confirmation.
Confirmation pattern incomplete. preview_bulk_task_update and preview_project_import return tokens, but no mechanism shown to prevent token reuse or expiry. Tokens could be replayed or stale.