Loading ToolBench
Lolik Trend Radar MCP server — live trend → money-ready intelligence as MCP tools for Claude and agents. Every tool call uses your Lolik API key (auth + credits + rate limits).
12 tools with consistent naming (verb_noun pattern) and functional descriptions. All tools are READ_ONLY, reducing complexity. However, input schemas are inferred from a declarative manifest rather than explicitly registered with the MCP SDK, this caps per-tool schema scores at 50. Parameter descriptions are present but terse (avg ~40 chars, below the 72-char baseline). Output schemas are not documented. Error handling is generic (ok/status/body structure) with no recovery guidance. No tool annotations (readOnlyHint, etc.) despite all being read-only.
Timing, momentum and lifecycle for a topic (rising vs peaking).
Daily what-to-post brief: phase-labeled pre-breakout alerts, measured outcome receipts, a validated pattern and the day's steepest mover. Drop-in replacement for a generic 'research trending topics' step in content pipelines.
Freshest news stories, optionally for a topic.
Top trending topics for a niche. Returns title, source, niche, score, volume.
Credits remaining + rate-limit status for your API key. Free (no credit cost).
Validated viral patterns (hook / structure) confirmed over time — what's working now, with evidence.
Input schemas inferred from declarative manifest, not explicitly registered with MCP SDK. Tool definitions are not directly visible in SDK registration code, only in tools.mjs TOOL_MANIFEST array. This violates the principle that tool definitions should be explicit and machine-readable at registration time.
Output schemas not documented. Tools return { ok, status, body } but the structure and fields of 'body' are not specified. LLMs cannot plan downstream calls or extract data reliably without knowing what fields to expect.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 56 | 2026-07-28+ | v2 |
Grade a draft title/hook against the patterns currently validated in live data — this week's data, not a static rubric. Score math runs in code; every component states its basis (measured / ai_judged / rule). Includes up to 2 rewrites with measured proof. 5 credits.
One-call content brief for a niche: what's trending, the best opportunities (with possible monetization), and timing. Combines several signals.
Ranked opportunities for a niche, with possible monetization paths (estimates, not guarantees).
What's accelerating now, before it peaks — early-mover signal for a niche.
Hooks for a topic, anchored in the title patterns currently validated across creators (live measured data, not a static list). Hook text is AI-drafted; the proof next to each hook (creators, days, strength) is measured. 5 credits.
Virality score + breakout probability + hook type for a topic or title.
Parameter descriptions are terse (avg ~40 chars vs 72-char baseline). E.g., 'Filter to a niche.' lacks context on valid values, use cases, or when to omit. 'Data source: google (default) or youtube.' mixes enum values into description instead of using formal enum constraint.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). All 12 tools are READ_ONLY but this is not declared in the schema. Agents cannot infer safety properties without explicit annotations.
Error handling is generic. Responses return { ok: false, status, body: { error, message } } but do not categorize errors as retryable, user-fixable, or fatal. No recovery guidance (e.g., 'Try search_users() first'). LLMs cannot decide next steps.