MCP server that dynamically provisions agent skills based on project context
Mother MCP Server has 15 tools with generally clear naming (verb_noun pattern) and descriptive documentation. However, there are significant gaps in schema completeness, parameter descriptions, and output schema documentation. Tools like 'setup', 'sync_skills', and 'preview_sync' have good narrative descriptions but lack detailed input parameter descriptions. Several tools are missing enum constraints (e.g., 'agent' parameter values are defined in schema but not always explained in descriptions). Output schemas are not documented for any tool, making it unclear to LLMs what fields to expect in responses. Error handling and recovery guidance are absent from tool descriptions. The server uses STDIO transport, which is a hard architectural limitation.
Check if any installed skills have newer versions available.
Confirm and apply skill changes from a preview. Use after preview_sync to apply approved changes. You can: - Apply all changes by just providing the preview_id - Selectively approve specific skills with approved_skills - Reject specific skills with rejected_skills
Get information about the detected agent (Claude or Copilot) and skill paths.
View the detected project context including tech stack, installed skills, and agent information.
Install a curated skill bundle by ID. This installs all skills in the bundle at once for a complete development stack. Use list_bundles to see available bundles and their IDs.
Manually install a skill from the registry by name.
Missing output schema documentation for all 15 tools. LLMs cannot determine what fields to expect in responses, forcing them to guess about return structure and preventing safe tool chaining.
Parameter descriptions are minimal or generic. 'auto_install' in 'setup' lacks detail about what 'automatic' means (automatic relative to what?). Parameters like 'force_refresh' and 'dry_run' in 'sync_skills' are explained but lack guidance on when LLMs should use them.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 53 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 48 | - | v1 |
List available skill bundles (curated collections for common use cases). Bundles group related skills together for quick setup of complete development stacks. Examples: "Full-Stack Next.js", "Python API", "DevOps Docker & CI/CD"
Preview skill changes without applying them. Shows what skills would be added, updated, or removed, distinguishing between: - **Manual skills**: Configured in always_include (user explicitly requested) - **Discovered skills**: Auto-detected from project tech stack - **Dependencies**: Required by other skills Returns a preview_id that can be used with confirm_sync to apply selected changes.
Re-scan project files to update detected technologies.
Reset Mother MCP by removing all installed skills and clearing configuration.
Search the registry for available skills by name, description, or tags.
Set the preferred agent (Claude, Copilot, or Codex) for skill installation.
**Start here!** Initialize Mother MCP for your project. This is the recommended first step when setting up Mother MCP. It will: 1. Scan your project to detect technologies (languages, frameworks, databases, tools) 2. Fetch the skill registry and find matching skills for your stack 3. Show recommended skills with match explanations 4. Let you choose which skills to install Use this tool when users say things like "setup mother mcp", "initialize", "get started", or "what skills do I need".
Synchronize project skills based on detected technologies. This tool: - Scans project files to detect your tech stack (package.json, requirements.txt, config files, etc.) - Compares with the skill registry to find matching skills - Shows a PREVIEW of changes and asks for user confirmation before applying - Distinguishes between manually configured skills (always_include) and auto-discovered skills - Downloads approved skills to the appropriate location (.github/skills, .claude/skills, or .codex/skills) IMPORTANT: This tool will show proposed changes and wait for user approval before making changes. Use dry_run=true to see changes without any prompt for confirmation.
Remove a skill from the project.
No error handling or recovery guidance in tool descriptions. Tools like 'install_skill' and 'uninstall_skill' do not explain what happens on failure (skill not found, permission denied, network error) or what the LLM should do next.
Destructive operations (reset_skills, uninstall_skill) lack confirmation/dry-run patterns. 'reset_skills' will delete all installed skills and optionally the config file, but the description does not mention a confirmation step or preview capability to prevent accidental data loss.
Missing enum constraints in descriptions. The 'agent' parameter accepts ['auto', 'claude', 'copilot', 'codex', 'both'] but descriptions do not explain when to use each or the implications of 'auto' vs explicit values.
Minimal descriptions for discovery/info tools. 'get_project_context' and 'get_agent_info' have single-sentence descriptions without explaining what structure they return, when to call them, or how their output feeds into other tools.