An in-process MCP server for AI-driven business operations management, providing tools for project tracking, task management, and approval workflows with role-based access control.
Server has 8 tools with basic schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. All tools have descriptions (10-80 chars), but most are terse and lack context for LLM selection. Parameter descriptions exist but are minimal. No output schemas documented. Error handling is present but generic. Tool naming follows verb_noun convention well (get_*, create_*). Security model includes approval gates for write operations, which is good. However, lack of pagination guidance, missing output field documentation, and sparse parameter constraints limit agent reasoning.
Create a new task in a project. Requires approval ID and approver role.
Retrieve blocked tasks, optionally filtered by project and assignee
Retrieve projects that are delayed, with optional limit parameter
Retrieve workload information for a specific developer
Retrieve overdue tasks, optionally filtered by project and assignee
Retrieve detailed information about a specific project
Retrieve notes for a specific project
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields without knowing response structure.
Pagination not documented. Tools returning lists (get_overdue_tasks, get_blocked_tasks, get_project_notes, search_client_messages) accept 'limit' but no offset/cursor or total_count guidance. Large result sets will blow context windows.
Tool descriptions are terse (50-70 chars). Missing WHEN to use, prerequisites, and what the tool returns. E.g., 'Retrieve overdue tasks' does not explain when to call this vs get_blocked_tasks or how results differ.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 43 | <=2025-11-25 | v2 |
Search client messages, optionally filtered by project and query string
Parameter descriptions lack format/constraint guidance. E.g., 'Filter by project ID' does not explain format (UUID? alphanumeric?). 'Maximum number of tasks to return' lacks min/max bounds. LLMs will guess invalid values.
Error responses are generic ('VALIDATION_ERROR', 'INTERNAL_ERROR'). No recovery guidance. LLM cannot determine if error is retryable, user-fixable, or fatal.