MCP server for KasGraph, a subgraph-style indexing system for Kaspa blockchain. Provides tools for querying indexed blockchain data, managing subgraphs, and inspecting indexing status.
KasGraph MCP server has 8 tools with consistent READ_ONLY semantics and basic parameter schemas. However, critical gaps undermine quality: (1) Tool names lack action verbs, 'committedBlock', 'poiCheckpoints', 'detectedPatterns' are noun-heavy and do not signal intent (pattern:tool requires verb_noun naming). (2) Descriptions are minimal (10-50 chars) and lack context for LLM selection, e.g., 'Query a committed block by subgraph and hash' does not explain WHEN to use this vs. committedBlocks, or what the response contains. (3) Parameter descriptions are present but terse; no constraints (enums, ranges, formats) documented. (4) Output schemas are completely absent, LLMs cannot plan downstream calls or extract chaining IDs. (5) No error handling guidance; no recovery paths. (6) No pagination limits stated despite 'first' parameters suggesting large result sets. Baseline: 90% of A+ tools have 50-200 char descriptions and documented output schemas; this server averages ~30 chars and zero output docs.
Query a committed block by subgraph and hash
Query committed blocks for a subgraph with pagination
Query the lineage history of a covenant UTXO, including genesis, transitions, and branches
Query detected covenant spends (block inputs that consumed tracked covenant UTXOs) with optional covenant ID filtering
Query detected patterns (covenant patterns, KRC-20, NFTs, etc.) with optional kind filtering
Query multiple mapping-emitted entities of a given type with pagination
Query a single mapping-emitted entity by subgraph, type, and ID
Query proof-of-indexing checkpoints with optional DAA score range filtering
Tool names lack action verbs. 'committedBlock', 'poiCheckpoints', 'detectedPatterns' are noun-heavy and do not signal intent. LLMs cannot infer action from name alone.
Descriptions are too short (10-50 chars) and lack context. Do not explain WHEN to use each tool, what distinguishes it from similar tools, or what the response contains. Baseline: A+ tools average 194 chars.
Output schemas are completely absent. LLMs cannot plan downstream calls, extract chaining IDs, or understand response structure. No documentation of what fields are returned.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 35 | 2026-07-28+ | v2 |
No pagination limits or guidance. Tools accept 'first' parameter but do not document max results, default behavior, or whether results are capped. Risk of context window exhaustion.
No error handling guidance. Tools do not document what errors are possible, whether they are retryable, or what the LLM should do next. No recovery paths.