Loading ToolBench
MCP server exposing startup engineering acceleration data, commit velocity, contributor growth, repository expansion across 350+ startup GitHub orgs in 15 sectors. Transparent, citable funding-likelihood scores with full evidence chains.
12 tools with complete schemas and descriptions, but quality varies significantly. Tool names follow verb_noun convention well (get_*, search_*, predict_*, compare_*). Descriptions range 50-250 chars, mostly adequate but some lack actionable context. All parameters have type definitions and descriptions. Output schemas are not explicitly documented in source. Error handling guidance is minimal. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite all being READ_ONLY. Missing recovery guidance for common failure modes (e.g., startup not found, API rate limits).
Compare engineering-acceleration signals across multiple startups side by side.
PAID per-request (€0.19/call via pre-paid credits or x402 USDC on Base; misses are FREE). Returns enriched signal: composite score, sector percentile, plain-English thesis, comparables, and multi-period history.
Public-source diligence dossier for a company or entity: M&A history, public backers, and the published engineering signal. Returns found:false with an honest note when outside the tracked corpus, never guesses.
Full methodology document covering data sources, signal classification, refresh cadence, and known limitations.
Compute a Scout Score (0-100) for a GitHub user from their starring history vs validated unicorns.
Period, sector and startup counts, last refresh, citation, and direct URLs to machine-readable formats.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite all 12 tools being READ_ONLY. Agents cannot distinguish safe-to-retry operations from those with side effects.
Output schemas not documented in source code. LLMs cannot plan downstream tool calls or extract required fields (e.g., startup IDs for chaining). Violates pattern:response-shaper.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 62 | 2026-07-28+ | v2 |
Full engineering-acceleration profile for a single startup, by display name or GitHub org slug.
Top 20 startups by engineering acceleration across all 15 sectors for the current weekly period.
Transparent, scored funding-likelihood claim for one tracked startup, with the full evidence chain, confidence level, honest caveats, and citable provenance (methodology + SSRN).
Every tracked startup within a sector, ranked by engineering acceleration.
Compose a ready-to-share social post (Twitter/Bluesky/Mastodon/LinkedIn/Telegram) about a prior tool result, with per-platform char counts and intent URLs. Composes only, never posts on the user's behalf (two-step approval).
Build and score a shortlist of N startups against a thesis, with optional sector and geography filters.
No error recovery guidance. When startup not found or API rate-limited, tools return errors without suggesting next steps (e.g., 'Try search_startups_by_sector() with a partial name'). Agents cannot self-correct.
Parameter 'sector' in search_startups_by_sector uses enum constraint in description (15 hardcoded sector slugs) but not as formal JSON Schema enum. LLMs may hallucinate invalid sector values.
share_result tool accepts 'result' as free-form string (prior tool output) and 'platform' as enum. No validation that result is valid JSON or structured output. Agents may pass malformed data.