A tool to create, build, and manage MCP servers for use with Claude and other MCP clients
The server defines 3 tools with reasonable schemas and descriptions, but exhibits significant gaps in error handling guidance, output schema documentation, and composition patterns. Tool names follow verb-noun convention (create-server, build-server, install-server), which is correct. Descriptions are present but generic (10-50 chars), falling below the 50-200 char baseline for LLM optimization. All parameters have type definitions and descriptions, but lack actionable error recovery guidance. The tools manipulate the filesystem and execute external commands (npm, build tools) without documented safety constraints, retry semantics, or compensation patterns. Output schemas are not formally documented, responses return unstructured text content rather than structured objects with typed fields. Error handling returns bare error messages without recovery hints (e.g., 'Failed to create server: <error message>' does not tell the LLM what to do next). Composition is reasonable (each tool has a single responsibility), but no pagination, batching, or multi-round-trip patterns are present.
Build an existing MCP server from source code
Create a new MCP server project with proper scaffolding
Install an MCP server for use with clients like Claude Desktop
Tool descriptions are too generic (10-50 chars). They state WHAT the tool does but lack guidance on WHEN to use it, dependencies, or expected outcomes. LLMs need actionable context to decide when to select a tool over alternatives.
Output schemas are not documented. Tools return unstructured text responses (type: text) rather than structured objects with typed fields. LLMs cannot extract specific values (e.g., the path where the server was created) for downstream tool calls without parsing prose.
Error messages do not guide recovery. Responses like 'Failed to create server: <error message>' tell the LLM an operation failed but not what to do next. Should include retryability, user-fixable fixes, or alternative tools to try.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 59 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 41 | - | v1 |
No confirmation pattern for destructive operations. build-server and install-server modify system state (write to filesystem, update Claude Desktop config) without dry-run or confirmation steps. An agent mistake could corrupt user configs.
Missing idempotency semantics. Calling create-server twice with the same name and directory will fail on the second call (directory already exists). No tool documents whether it is safe to retry or what happens on repeated calls.
No input validation constraints. Parameters accept arbitrary strings; there is no regex pattern, length limit, or character restriction documented. 'name' parameter should reject names with special characters or spaces that would break project scaffolding.
Parameter descriptions lack format/constraint details. 'directory' is described as 'Directory where the server should be created' but does not specify: must be absolute path? relative to cwd? must exist or will be created? These gaps force LLMs to guess.
Helper functions (createTypeScriptProject, buildTypeScriptProject, etc.) are referenced but implementation is incomplete in the source excerpt. Cannot verify that scaffolding actually produces valid MCP servers or that build steps are correct.