DevAlpha MCP Server — AI-native on-chain intelligence for builder agents. Connects AI agents to DevAlpha's ecosystem intelligence, opportunity discovery, and developer activity data across 50+ ecosystems (blockchains and AI platforms).
DevAlpha MCP Server provides 3 tools with reasonably clear purposes and basic schemas. However, critical gaps exist: (1) Output schemas are NOT documented, tools return JSON strings via _json_response() helper, but the response structure is not formally declared; (2) Parameter descriptions are present but lack constraint details (enums, ranges, format specs); (3) No error handling guidance in descriptions; (4) No tool annotations (readOnlyHint, etc.) despite all being READ_ONLY operations. The server is functional but falls short of production-grade tool design. Average per-tool score: 62.
Compare blockchains by key metrics. Returns a ranked list. Use this to answer: "Which chains have the most TVL?", "Best L2s by developer activity?", "Compare Solana vs Base vs Arbitrum".
Get detailed metrics for a specific blockchain. Use this to answer: "Tell me about Solana", "What's Ethereum's TVL?", "How active are developers on Base?".
Search hackathons, bounties, grants, and demand signals. Use this to answer: "Active hackathons on Solana", "DeFi bounties", "What do developers want to build?", "Grants available now".
Output schemas not documented. Tools return JSON strings without formal response type declarations. LLMs cannot plan downstream calls or extract fields without knowing what fields exist in the response.
Parameter descriptions lack constraint specifications. 'sort_by' accepts 'tvl, fees, dex_volume, commits, protocols, name' but this is documented in prose, not as an enum constraint. Same for 'chain_type' (l1, l2), 'opportunity_type' (hackathon, bounty, grant, pain_point, tool_request, market_signal). LLMs cannot reliably select from prose-listed options; enums are machine-parseable.
No tool annotations present. All three tools are READ_ONLY (no side effects), but MCP tool definitions do not include readOnlyHint annotations. This prevents clients from optimizing caching, retries, and UI presentation.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 57 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 25 | - | v1 |
No error recovery guidance in descriptions. Tools call external REST API but descriptions do not hint at error cases or next steps (e.g., 'If chain not found, try search_opportunities to find related data'). Pattern: recovery-guide.
Parameter type annotations use Python union syntax ('str | None') which is not portable to JSON Schema. The input schema should be explicit JSON Schema with 'type' and 'oneOf' constructs for clarity across platforms.