Loading ToolBench

Reading the current corpus…

fastmcp-builder-template MCP server · D · ToolBench

← Browse

fastmcp-builder-template

github.com/paulieb89/fastmcp-builder-template

Local-first FastMCP v3 builder scaffold for designing and reviewing MCP servers.

2026-07-28+STDIO16 tools

57/100

Definition Quality76

Protocol Readiness48

Supportability31

Spec posture

Inferred effective spec 2026-07-28+·current baseline 2026-07-28

Static source inference · medium confidence

Current-spec patterns detected

Summary

FastMCP Builder is a meta-tool server for reviewing and scaffolding MCP servers. 16 tools with mostly complete schemas and descriptions. Naming is clear and verb-forward (classify_, review_, extract_, check_, suggest_, generate_). Descriptions range 50 - 250 chars, meeting the 10 - 1024 baseline. However, parameter descriptions are sparse or generic ('use_case', 'actor', 'expected_control_pattern' lack specificity). Output schemas are inferred from Pydantic models but not explicitly documented in tool definitions. Error handling is minimal, tools return structured objects but lack recovery guidance. Tool annotations are declared in the server but not consistently applied across all tools.

Tools (16)

check_error_response_designread onlysource verified77/100

Classify tool failure modes into practical error-response categories.

check_prompt_argument_descriptionsread onlysource verified77/100

AST-scan a FastMCP server source for `@mcp.prompt` arguments that lack descriptions. Descriptions can live in `Annotated[X, Field(description=...)]` or in the docstring's `Args:` block. Spec source: MCP — `PromptArgument.description` SHOULD be set. Severity MEDIUM (SHOULD, not MUST).

check_prompt_name_formatread onlysource verified73/100

Check whether a proposed FastMCP prompt name follows naming conventions.

check_resource_mime_type_declaredread onlysource verified78/100

AST-scan a FastMCP server source for `@mcp.resource` decorators that don't declare a `mime_type=` kwarg. Spec source: FastMCP — `servers/resources.md`. Severity MEDIUM (FastMCP-recommended, not MCP-mandatory). Without `mime_type`, FastMCP infers from the return type — fine for plain strings, brittle otherwise. Findings carry `spec_source="FastMCP"` and `spec_section="servers/resources.md#mime_type"`.

check_silent_error_returnsread onlysource verified82/100

Scan a FastMCP server's Python source for the silent-failure-conversion anti-pattern: an @mcp.tool function that returns an error-shaped sentinel (a dict with an "error" key, or a string starting with "Error") instead of raising a structured error. Catches both direct patterns (`return {"error": ...}` in the tool body) and one level of indirection (`return _handle_error(e)` where the helper itself returns error sentinels — the bailii-mcp pattern). Does not execute the source. Returns a SilentErrorReport whose `findings` are ReviewFinding objects (severity=HIGH, code="tool.silent_error_return"). Same shape as review_fastmcp_manifest so the design-review skill can merge findings uniformly by severity.

Issues (5)

HIGHtool-description ↗

Parameter descriptions are generic or missing context. 'use_case', 'actor', 'expected_control_pattern', 'manifest', 'path', 'name', 'uri_pattern' lack specificity about expected format, constraints, or examples. LLMs cannot infer whether 'path' is a file path, module path, or URI.

classify_mcp_primitivecheck_tool_annotations_declaredcheck_prompt_argument_descriptionscheck_resource_mime_type_declaredcheck_silent_error_returnscheck_prompt_name_formatcheck_tool_name_formatcheck_uri_stability

HIGHresponse-shaper ↗

Output schemas are not explicitly documented in tool definitions. Tools return Pydantic models (PrimitiveClassification, ManifestReview, SilentErrorReport, etc.) but the schema structure is inferred from imports, not visible in the tool registration. LLMs cannot see what fields to expect.

classify_mcp_primitivereview_fastmcp_manifestextract_manifest_from_sourcecheck_tool_annotations_declaredcheck_prompt_argument_descriptionscheck_resource_mime_type_declared

Recommendations

Add explicit parameter descriptions to all tools. For 'path', specify: 'Absolute or relative file path to a Python module containing FastMCP server definitions (e.g., /home/user/my_server.py or ./src/server.py).' For 'manifest', specify: 'A dict with keys: name (str), tools (list of dicts with name, description, input schema), resources (list), prompts (list).'
Document output schemas inline in tool descriptions or via a separate schema field. E.g., 'Returns a PrimitiveClassification object with fields: recommendation (str: TOOL|RESOURCE|PROMPT), rationale (str), confidence (float 0 - 1), alternatives (list[str]).'
Add error handling guidance to tools that parse or validate input. E.g., extract_manifest_from_source: 'If the path does not exist or is not a valid Python file, raises FileNotFoundError. If the module has no FastMCP server, returns an empty manifest.'
Apply tool annotations to all 16 tools in src/fastmcp_builder/server.py. All are read-only and idempotent; add annotations={'readOnlyHint': True, 'idempotentHint': True, 'destructiveHint': False, 'openWorldHint': False} to each @mcp.tool decorator.
Migrate from STDIO to HTTP+SSE or Streamable HTTP transport to enable remote client access and improve protocol readiness. Update FastMCP initialization to use HTTP listener instead of stdio().

Score history

Overall score trend

First recorded score · v2 rubric

57/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	D	57	2026-07-28+	v2