hubmeshSAFE
Centrality-aware GraphRAG retrieval planner — drop-in layer over any vector DB. Zero LLM in the query path; MCP server included.
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
[](https://github.com/DemigodDSK/hubmesh/actions/workflows/test.yml) [](https://github.com/DemigodDSK/hubmesh) [](https://github.com/DemigodDSK/hubmesh/blob/main/LICENSE) [](https://github.com/DemigodDSK/hubmesh/releases)
Centrality-aware GraphRAG retrieval planner. Drop-in layer over any vector DB.
hubmesh is a Python library that improves multi-hop RAG quality on top of an existing vector database. You don't replace your infrastructure — you add a smart planner between your vector DB and your LLM.
What problem this solves
Naive vector retrieval ("embed query, get top-k by cosine similarity") fails on multi-hop questions like "Where was the founder of the company that acquired Slack born?" The correct answer requires retrieving entities along a reasoning path, not the single most similar item.
GraphRAG and HippoRAG showed that running a small Personalized PageRank over a knowledge graph at query time can substantially improve multi-hop retrieval. hubmesh extends that line with two contributions:
- Entity-anchored seeding, multi-component ranking. In KG mode the PPR
seeds are the question's own entities resolved against the corpus graph (alias index; falls back to the entities of the top cosine matches when the question names none); in kNN mode they are the ANN top-k. The multi-component score — cosine relevance, pooled PPR mass, and multi-anchor convergence, min-max normalized and fused 3:1:1 — is applied to the document ranking, not to seed choice.
- Budget-aware context packing. Once relevant entities are scored, pack them into
the LLM
a0ef1290c8f2OBSERVED · 2026-10-11Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.
claude mcp add hubmesh -- None hubmesh==0.4.2
Exposed tools (8)
8 read · 0 write · 0 destructive.
| Tool | Risk | Description |
|---|---|---|
entity_neighbors | read | Explore around an entity: which documents mention it and which |
get_document | read | Full text of one document (retrieve() returns snippets only). |
graph_stats | read | Corpus overview: sizes plus the highest-degree entities — the |
index_corpus | read | Index documents into a named corpus: embeds them, builds the |
list_corpora | read | List indexed corpora with their stats (doc / node / edge counts). |
path_between | read | Connection paths between two entities through the entity-document |
resolve_entities | read | Resolve free-text entity names to knowledge-graph nodes (alias |
retrieve | read | Graph-aware retrieval: entity seeds + Personalized PageRank over |
Trust audit
SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | PASS |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (3 observation(s))
- Network
- none-observed
- Shell
- none-observed
- Dependencies
- pinned
- Secrets in source
- none-found
Findings (9)
complementarity_hotpotqa_n500_bge-m3.json.gz
complementarity_hotpotqa_n500_minilm.json.gz
complementarity_musique_n300_bge-m3.json.gz
complementarity_musique_n300_minilm.json.gz
m = __import__(mod)
return f"{title}::{hashlib.sha1(text.encode('utf-8')).hexdigest()[:8]}"benchmarks/results/ablation_coherence_musique_n2417.json
benchmarks/results/complementarity_2wiki_n1000_bge-m3.json.gz
benchmarks/results/complementarity_2wiki_n1000_minilm.json.gz
Gates applied: no_behavioural_pass.
a0ef1290c8f2full audit observations/trust-audit/mcp-server/demigoddsk__hubmesh.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-11 | a0ef1290c8f2 | SAFE | B | 89 | first audit |
Questions
What is the hubmesh MCP server?
Centrality-aware GraphRAG retrieval planner — drop-in layer over any vector DB. Zero LLM in the query path; MCP server included.
What tools does hubmesh expose?
8 in total: 8 read-only, 0 that write, and 0 that can delete or overwrite. Every one is listed on this page with its risk.
Is hubmesh safe to connect to an agent?
The audit found nothing in the source that contradicts what it says it does, and graded it B (89/100). Grade A is held back for packages that have also passed a sandboxed behavioural run, which is why a clean server reads B.
What credentials does hubmesh need?
It reads HUBMESH_API_KEY from the environment. Give it a token scoped to the least it needs — an agent that can be talked into calling a tool can be talked into calling it with your credentials.
How does hubmesh run?
It speaks stdio, so it runs as a local process your client starts. It is published on PyPI as hubmesh.
How current is this page?
The grade is for one exact copy of the source (a0ef1290c8f2), read on 2026-10-11. The repository is watched and re-audited when it changes.