Loading ToolBench

Reading the current corpus…

openrouter_task2model MCP server · C · ToolBench

← Browse

openrouter_task2model

github.com/kaminoguo/openrouter_task2model

MCP Server to find the best AI model for your task. Searches 300+ OpenRouter models by semantic match and constraints.

<=2025-11-25STDIO3 tools

62/100

Definition Quality70

Protocol Readiness42

SupportabilityNot measured-

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

Three tools with clear semantic purpose and actionable descriptions, but inconsistent schema completeness and missing output documentation. All tools have descriptions (10 - 20 chars minimum exceeded), but the task2model tool lacks explicit output schema documentation despite having complex nested input. Parameter descriptions are present but sometimes generic. Error handling returns structured errors but lacks recovery guidance for LLMs.

Tools (3)

get_model_profileread onlysource verified73/100

Get detailed profile of a specific model, optionally including endpoints and parameters.

sync_catalogread onlysource verified73/100

Refresh the OpenRouter models catalog cache. Use force=true to bypass TTL.

task2modelread onlyauthsource verified63/100

Recommend models for a task. Returns top 100 model IDs filtered by age (<1 year) and ranked by semantic match. IMPORTANT: Use default parameters (just provide "task"). Only add constraints if user explicitly requests them. Defaults: limit=100, detail=names_only, max_age_days=365 Optional: hard_constraints.max_price_per_1m, hard_constraints.required_parameters, result.detail (minimal|standard|full)

Issues (6)

HIGHdocumented-output-schema ↗

Missing output schema documentation for all three tools. LLMs cannot infer what fields to extract from responses.

sync_catalogget_model_profiletask2model

HIGHconstrained-input ↗

task2model input schema lacks clarity on enum meanings: 'routing' enum values (price, throughput, latency) and 'detail' enum values (names_only, minimal, standard, full) are undocumented. LLMs will guess what each means.

task2model

HIGHpaginated-result ↗

task2model does not document pagination. Description claims 'Returns top 100 model IDs' but no limit parameter is visible in the input schema (limit is nested in result object with default=100). No next_cursor or offset documented for handling >100 results.

task2model

MEDIUMrecovery-guide ↗

Error responses (e.g., 'INVALID_INPUT', 'Model not found') are structured but lack recovery guidance. When get_model_profile returns 'Model not found: X', the description should suggest 'Try sync_catalog() to refresh the catalog' or 'Call search_models() with a partial name'.

Recommendations

Document the output schema for each tool. Add JSDoc or TypeScript comments describing SyncCatalogResult, ModelProfileResult, and task2model result structure (e.g., '@returns {object} result - {model_count: number, fetched_at: string, source: string, auth_used: boolean}'). Include this in the tool description or in a separate 'outputSchema' field.
Add descriptions to enum values. For task2model.preferences.routing, explain: 'price: rank models by cost-per-token; throughput: rank by input/output tokens per second; latency: rank by response time.' For result.detail, explain: 'names_only: return [id1, id2, ...]; minimal: return [{id, price}]; standard: return full model objects; full: include endpoints and parameters.'
Clarify pagination handling. If task2model always returns ≤100 results, state this explicitly: 'Returns up to 100 model IDs (limit parameter default=100, max=500). Use limit parameter to request fewer. If >500 total matches, no pagination: agent must refine constraints.'
Add recovery guidance to error responses. Modify error handling: instead of just returning {'code': 'INVALID_INPUT', 'message': 'Invalid TaskSpec'}, return {'code': 'INVALID_INPUT', 'message': '...', 'nextSteps': 'Ensure task is a non-empty string. If you provided hard_constraints, verify that required_parameters, input_modalities, output_modalities, and providers are valid arrays.'}'.
Remove implementation-level guidance from tool descriptions. Rewrite task2model description to: 'Recommend AI models for a task. Searches 300+ OpenRouter models by semantic match and returns top matches filtered by age and constraints. Default: 100 newest models, ranked by semantic relevance. Optionally filter by price, required parameters, and modalities.' Move 'Use default parameters' guidance to agent instructions.

Spec posture evidence

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

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

Score history

Overall score trend

↑ 62 points across a rubric change (v1 → v2)

62/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-22	C	62	<=2025-11-25	v2
2026-03-09	F	0	-	v1