hubmesh MCP server
Deterministic multi-hop graph retrieval for RAG. Zero LLM calls in the query path.
2 stars130 downloads/wk
Reviews
Write oneNobody has reviewed hubmesh yet.
If you have run it, two minutes of your experience saves the next person an afternoon.
hubmesh tools (8)
write = sends, deletes, buys or postsRead from the package source without running it. The installed server may list more.
entity_neighborsExplore around an entity: which documents mention it and which entities it connects to (with co-occurrence weight / predicates). Useful for choosing the next hop when retrieve() alone is ambiguous.
get_documentFull text of one document (retrieve() returns snippets only).
graph_statsCorpus overview: sizes plus the highest-degree entities — the hubs PPR diffusion flows through.
index_corpusIndex documents into a named corpus: embeds them, builds the entity knowledge graph (spaCy NER), and persists everything to disk. `documents` is a list of {"id": str, "text": str}. Re-using an existing name replaces that corpus. Indexing cost is paid once — queries afterwards are ~milliseconds.
list_corporaList indexed corpora with their stats (doc / node / edge counts). Start here to see what is queryable.
path_betweenConnection paths between two entities through the entity-document graph — up to `k_paths` paths, shortest first. INTERPRETATION CAUTION: the shortest path can ride an incidental co-mention (two names in the same sentence); longer paths with more `via_documents` often reflect the more meaningful chain. Report the nature of the connection, not just its existence.
resolve_entitiesResolve free-text entity names to knowledge-graph nodes (alias index + fuzzy fallback). Use before seeding retrieve() to confirm an entity exists and see its canonical label.
retrieveGraph-aware retrieval: entity seeds + Personalized PageRank over the corpus knowledge graph, fused with cosine similarity.
Public scan report
scanner v0.1.5 · 2026-09-19 · same rubric, same numbers if you re-run it
- Code scan20 source files scanned25/25
- –Live reliabilityno gateway calls yet and no remote to proben/a
- –Tool poisoningtools not inspected (local package is not executed); not countedn/a
- Auth qualitylocal package, no credentials required12/15
- Maintenancelast push 1 days ago15/15
- Maintainer identityregistry namespace matches repository owner; GitHub account older than a year8/10
Install directly
claude mcp add hubmesh -- uvx hubmesh
hubmesh: common questions
- Is hubmesh MCP server safe?
- Yes, by our scan: it is graded A (92/100). Read the hubmesh safety report
- How do I install hubmesh?
- It runs on your machine. Copy the Claude Code, Claude Desktop or Cursor config from the install section.
- Does hubmesh need an API key?
- Not as far as the registry entry and our scan can tell: no credentials are declared or required.
- Is hubmesh maintained?
- The last commit was 2 days ago (2026-09-19). The latest release is v0.4.2.
- What can I use instead of hubmesh?
- Servers from other publishers that do the same job: Google Surf MCP server, Tw Legal Rag MCP server and haiku.rag MCP server. Compare all hubmesh alternatives.
Alternatives to hubmesh
Same job from other publishers: the closest match first, then the best rated.
- Google SurfWeb, academic and code search with graph RAG, data lineage, ontology and cross-project schema links.not reviewedEstablishedB
Tw Legal RagCitation-guarded retrieval over 22M Taiwan court judgments and administrative interpretationsnot reviewedEstablishedA- haiku.ragLocal-first agentic RAG with citations - hybrid search, reranking, multimodal document retrievalnot reviewedEstablishedA
- adrkit decision memoryDeterministic, offline, read-only ADR decision memory for coding agents. No model or network calls.not reviewedGrowingA
- rag-ratRepository intelligence, code graph, history, papertrail, and cross-agent memory for coding agents.not reviewedGrowingB