Loading ToolBench

Reading the current corpus…

finch MCP server · F · ToolBench

finch

A Dart-based web application framework with MCP (Model Context Protocol) server capabilities for routing, database connectivity, WebSocket management, and scheduled tasks

<=2025-11-25STREAMABLE_HTTP10 tools

36/100

Definition Quality22

Protocol Readiness45

Supportability53

Spec posture

Inferred effective spec <=2025-11-25·current baseline 2026-07-28

Static source inference · medium confidence · detected: Logging

Deprecated protocol patterns detected

Summary

Finch MCP server has critical definition quality gaps. Tool names lack action verbs (tools/list, resources/read, prompts/get use path-style naming rather than verb_noun convention). Descriptions are present but minimal (10-50 chars), below the 50-200 char baseline for LLM optimization. Input schemas visible for tools/call, resources/read, prompts/get, and logging/setLevel, but most lack parameter descriptions. tools/list, resources/list, resources/templates/list, prompts/list, initialize, and server/discover have no visible input schemas. Output schemas are undocumented across all tools. Error handling and recovery guidance absent. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk classifications in metadata.

Tools (10)

initializeread onlysource verified15/100

Initializes the MCP server connection and returns server capabilities

logging/setLevelwrite42/100

Sets the logging level for the MCP server

prompts/getread only40/100

Retrieves a specific prompt by name with optional arguments

prompts/listread only20/100

Lists all available prompts registered with the MCP server

resources/listread only20/100

Lists all available resources registered with the MCP server

resources/readread only43/100

Reads the contents of a specified resource

resources/templates/listread only20/100

Lists all available resource templates registered with the MCP server

server/discoverread onlysource verified22/100

Discovers server capabilities and supported protocol versions (MCP 2026-07-28)

Issues (5)

CRITICALtool-naming ↗

Tool names use path-style (tools/list, resources/read) instead of verb_noun convention (list_tools, read_resource). LLMs cannot infer action intent from path names; verb_noun is the production standard across 549 baseline tools.

tools/listtools/callresources/listresources/readresources/templates/listprompts/listprompts/getserver/discover

CRITICALconstrained-input ↗

6 of 10 tools have no visible input schemas (tools/list, resources/list, resources/templates/list, prompts/list, initialize, server/discover). LLMs cannot validate parameters without schemas.

tools/listresources/listresources/templates/listprompts/listinitializeserver/discover

HIGHtool-description ↗

All tool descriptions are 10-50 characters, well below the 50-200 char baseline for LLM optimization. Descriptions lack WHAT/WHEN/WHY context. E.g., 'Lists all available tools' does not explain when to call it vs. server/discover, or what structure is returned.

Recommendations

Rename all tools to verb_noun convention: tools/list → list_tools, resources/read → read_resource, prompts/get → get_prompt, initialize → initialize_server, server/discover → discover_server_capabilities, logging/setLevel → set_logging_level. This matches production baselines and enables LLMs to infer intent from names alone.
Expand all descriptions to 50-200 characters with WHAT/WHEN/WHY context. Example: 'list_tools: Returns all registered tools with names, descriptions, and input schemas. Call this first to discover available capabilities before invoking tools/call.' Current 'Lists all available tools' is too vague.
Add input schemas to tools/list, resources/list, resources/templates/list, prompts/list, initialize, and server/discover. Even discovery tools should declare their parameters (e.g., does list_tools accept a filter parameter?). Use JSON Schema with type and description for each parameter.
Document output schemas for all 10 tools. Specify field names, types, and whether results are paginated. Example: 'tools/list returns {tools: [{name: string, description: string, inputSchema: object}], total: number}'. This enables LLMs to chain calls correctly.
Add tool annotations to the MCP protocol response: mark tools/list, resources/list, prompts/list, server/discover with readOnlyHint=true; mark tools/call and logging/setLevel with destructiveHint=true. This signals to agents which operations are safe to retry.

Spec posture evidence

Inferred effective spec: <=2025-11-25.

Relies on Logging (deprecated) - log to stderr or use OpenTelemetry

Score history

Overall score trend

First recorded score · v2 rubric

36/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	F	36	<=2025-11-25	v2