Skip to main content
Glama
AntoniovanDijck

MeshMind

by AntoniovanDijck
OverviewSchemaSourceRelated ServersDiscussionsAdmin
TypeScript
Local

Server Configuration

Describes the environment variables required to run the server.

NameRequiredDescriptionDefault
MESHMIND_HOMENoDirectory for persistent cache (default ~/.meshmind)
MESHMIND_CACHE_MAXNoMaximum number of cached entries (default 500)

Instructions

Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.

This server publishes no instructions, or was last inspected before Glama recorded them.

Capabilities

Features and capabilities supported by this server

Protocol revision2025-11-25

CapabilityDetails
tools
{
  "listChanged": true
}

Tools

Functions exposed to the LLM to take actions

NameDescription
scan_local_codebaseA

Recursively scan a directory and build a dependency graph: files, top-level symbols, import edges, an inferred call graph (with EXTRACTED/INFERRED/AMBIGUOUS confidence), community clustering, and structural analysis (hub/god nodes, import cycles, orphans). Default returns a compact summary; raw=true returns the full JSON map.

export_codebase_graphA

Scan a directory and export its dependency graph as either a Mermaid diagram (format='mermaid') or a D3/Obsidian-friendly nodes+edges JSON (format='json').

research_last_30_daysA

Fetch recent community/social signal on a topic from keyless public sources (Hacker News + comment enrichment, Reddit w/ RSS fallback, GitHub, Web, Lobsters, Bluesky, Stack Overflow, Lemmy), filtered to a trailing window. Results are relevance-reranked and deduped. Set compress=true to pipe the result through the token-reduction pipeline.

get_optimized_contextA

Reversible token-reduction pipeline. Accepts raw text OR a filePath. Three modes of use: (1) set targetTokens and MeshMind auto-escalates algorithms until the output fits your budget; (2) set explicit algorithms; (3) leave both for sensible defaults. Algorithms: strip, whitespace, line-dedup, json-min, truncate, stopwords, summarize. Set summarize=true for an abstractive summary via the host LLM (MCP sampling), extractive fallback. preview=true shows the per-step savings WITHOUT storing a ref. Returns compressed payload, exact BPE savings, and a ref for retrieve_context. Provide exactly one of text or filePath.

crush_fileA

Shortcut for the common 'this file is too big to read' case: reads a local file and compresses it in one call. With targetTokens, auto-escalates the pipeline until it fits; otherwise applies the default lossless-ish pipeline. Returns the compressed payload, exact BPE savings, and a reversible ref.

retrieve_contextA

Recover the original, uncompressed text for a ref returned by a prior get_optimized_context or compressed research call (reversible / CCR).

context_statsA

Return token-savings stats: session (this process) and lifetime (persisted across restarts under MESHMIND_HOME) — compress calls, original vs. crushed tokens, percent saved, cached refs, first/last seen.

Prompts

Interactive templates invoked by user choice

NameDescription

No prompts

Resources

Contextual data attached and managed by the client

NameDescription

No resources

Deploy Server
One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

TDQS

A4/5.0

Scored across 7 tools

Disambiguation4/5

Tools have largely distinct purposes, with some overlap between crush_file and get_optimized_context as both compress, but they target different use cases (file shortcut vs. general pipeline). export_codebase_graph and scan_local_codebase are related but differ in output format and depth. Overall, agents can distinguish most tools.

Naming Consistency5/5

All tool names use snake_case with a consistent verb_noun or descriptive pattern (e.g., crush_file, get_optimized_context, research_last_30_days). No mixing of conventions, making naming predictable and clear.

Tool Count5/5

Seven tools cover the advertised functionality—context compression, codebase analysis, and research—without being excessive. Each tool has a defined role, and the count is appropriate for the server's purpose.

Completeness4/5

The tool set covers key operations: compression, decompression, stats, codebase scanning, graph export, and research. Minor gaps exist, such as no tool to list or manage stored refs, and no direct configuration update, but core workflows are well-supported.

Maintenance

ActivityInactive
ResponsivenessNo issues

Resources