CLI tool for managing Gyazo image captures with MCP server support for AI agents
The server exposes 16 tools with varying quality. Most tools have descriptions and parameter schemas, but they are shallow and lack LLM optimization. Tool names are action-oriented and clear (list, get, search, collection, config, apps, domains, tags, locations, summary, stats, upload, sync, import, triage, touch), which is good. However, descriptions are mostly one-liners without context on when to use each tool or how they relate. Parameters lack descriptions in several tools (triage and touch have no input schema visible). Output schemas are not documented, LLMs cannot infer what fields to expect from responses. Error handling is minimal. The server operates over STDIO only, which caps it at 50 maximum for protocol readiness. Overall, this is a functional but mediocre implementation typical of community CLI tools adapted to MCP without deep agent optimization.
Rank metadata app names for a specific date
Show a collection and the images in it
Manage configuration
Rank metadata URL domains for a specific date
Get detailed metadata for an image
Import legacy data (type: json|hourly)
List recent images
Two tools (triage, touch) have no visible input schema or parameter descriptions. Their input definitions are shown as empty ({}) in the spec. This makes them unusable by agents, LLMs cannot infer what parameters to pass or what they control.
Tool descriptions are universally terse and lack context. Examples: 'List recent images', 'Search images', 'Get detailed metadata for an image'. These 1 - 5 word descriptions do not explain when to use each tool (when would you use 'list' vs 'search'?), what format the output is in, or how they compose. The rubric baseline is 194 chars average; these average ~30 chars.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 52 | 2026-07-28+ | v2 |
Rank metadata locations for a specific date
Search images
Show weekly stats summary in Markdown
Show weekly summary with daily uploads and metadata rankings
Sync images from yesterday back to N days
Rank metadata tags for a specific date
Touch and update image metadata
Triage and manage images
Upload an image file to Gyazo
Parameter descriptions are minimal or missing. For example, in 'list', the 'page' param has description 'page number', LLMs need to know what happens if you pass page 999 (is it clamped? Does it error? Are there 999 pages?). The 'hour' param format 'yyyy-mm-dd-hh format' is stated but not validated. The rubric baseline is 72 chars average per param; most here are under 20.
No output schemas are documented anywhere. For example, what does 'list' return? Is it a JSON array of image objects? What fields does each object have? Does it include pagination metadata (total_count, next_page)? LLMs cannot plan downstream calls without knowing the structure. The rubric requires 100% of A+ tools to have documented return types.
The 'config' tool modifies state (sets credentials) but the description does not acknowledge this or explain the consequences. Similarly, 'upload', 'sync', 'import' are marked WRITE but their descriptions do not state they modify state. The rubric requires descriptions of state-modifying tools to say so explicitly. This matters for agents deciding whether to retry or plan side effects.
No error handling is documented. What happens if an image ID is invalid? If the API rate-limits? If a file does not exist? LLMs receive raw errors with no guidance on recovery. The rubric requires error responses to tell the LLM what to do next and categorize errors as retryable, user-fixable, or fatal.
Parameter 'cache' is present in many tools (list, get, search, apps, domains, tags, locations, summary, stats, sync) with default true, but the description does not explain what cache means (local file cache? Redis? API-side cache?) or when it should be disabled. LLMs cannot reason about trade-offs without this context.
The 'date' parameter appears in apps, domains, tags, locations, summary, stats with format 'yyyy|yyyy-mm|yyyy-mm-dd' but no validation documented. Can you pass '2024-13-40'? What does 'yyyy' alone mean (current year? Any year?)? Does the API clamp invalid dates or error? Without constraints, LLMs will pass invalid values.
Mutually exclusive parameters are not documented. In 'list', both 'hour' and 'date' filter by time, can you pass both? What happens if you do? In 'apps', 'today' is documented to override 'date', but this is not stated in the 'date' parameter description itself. The rubric requires documenting mutually exclusive parameters in both descriptions.