mcpField
M
mcp-mat

mcp-mat

Local
github.com/codelipenghuiDeveloper Tools28Updated 32w ago

About mcp-mat

mcp-mat is a headless MCP server for Eclipse MAT that analyzes Java heap dumps by running the MAT launcher directly with java -jar. It is aimed at developers who need to investigate Java memory problems from an AI assistant without the MAT graphical interface.

Key features

  1. 01Predefined MAT reports such as leak suspects
  2. 02OQL queries returning inline results
  3. 0356 built-in MAT analysis commands
  4. 04Index artifact status check for a heap dump
  5. 05OQL parser guidance and supported patterns

Use cases

  1. 01Investigating memory leak suspects in a Java heap dump
  2. 02Reviewing dominator trees and paths to GC roots
  3. 03Checking thread overviews and thread stacks

MCP support

This is an MCP server that exposes MAT health checks, report runs, OQL queries and analysis commands as tools for MCP clients.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

mcp.json · Universal config
{
"mcpServers": {
"codelipenghui-mcp-mat": {
"command": "npx",
"args": [
"-y",
"mcp-mat"
],
"env": {
"MAT_HOME": "<MAT_HOME>",
"JAVA_PATH": "<JAVA_PATH>",
"MAT_XMX_MB": "<MAT_XMX_MB>"
}
}
}
}
Run commandnpx -y mcp-mat
Environment variables to set
MAT_HOMEPath to Eclipse MAT installation directory (e.g., /Applications/MemoryAnalyzer.app/Contents/Eclipse)
JAVA_PATHPath to Java executable (default: java)
MAT_XMX_MBMaximum JVM heap size in MB for MAT (default: 4096)

Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.

Packages
npmmcp-mat0.1.0

Tools

6

Saved tool list · updated 2w ago

ParameterTypeRequiredDescription
mat_homestringNo
java_pathstringNo

Details

From the project README

mcp-mat

Headless MCP server for Eclipse MAT using direct java -jar org.eclipse.equinox.launcher_*.jar execution.

Features

  • mat_healthcheck — Validate MAT launcher and Java runtime availability
  • mat_parse_report — Run predefined MAT reports (leak suspects, system overview, etc.)
  • mat_oql_query — Execute OQL queries and return inline results
  • mat_run_command — Execute 56 built-in MAT analysis commands (histogram, dominator_tree, path2gc, thread_overview, etc.)
  • mat_index_status — Check whether MAT index artifacts exist for a heap dump
  • mat_oql_spec — Return OQL parser guidance and supported patterns

Supported Commands (mat_run_command)

Dominator tree analysis

dominator_tree, show_dominator_tree, immediate_dominators, big_drops_in_dominator_tree

Path to GC roots

path2gc, merge_shortest_paths, gc_roots

Histogram & object listing

histogram, delta_histogram*, list_objects, group_by_value, duplicate_classes

Leak detection

leakhunter, leakhunter2, find_leaks, find_leaks2, reference_leak

Thread analysis

thread_overview, thread_details, thread_stack

Collection analysis

collection_fill_ratio, collections_grouped_by_size, array_fill_ratio, arrays_grouped_by_size, hash_entries, map_collision_ratio, extract_list_values, hash_set_values, primitive_arrays_with_a_constant_value

Reference analysis

references_statistics, weak_references_statistics, soft_references_statistics, phantom_references_statistics, finalizer_references_statistics

Finalizer analysis

finalizer_overview, finalizer_thread, finalizer_queue, finalizer_in_processing, finalizer_thread_locals

Retained set

show_retained_set, customized_retained_set

Component & top consumers

component_report, component_report_top, top_consumers, top_consumers_html, pie_biggest_objects

String & memory waste

find_strings, waste_in_char_arrays

Heap info & misc

heap_dump_overview, unreachable_objects, system_properties, class_references, comparison_report*

Eclipse/OSGi specific

bundle_registry, leaking_bundles

Export

export_hprof

* Requires a baseline heap dump (second .hprof file).

OQL mode notes

  • This server normalizes client OQL input and wraps it for MAT parse-app command mode.
  • You can send OQL with or without outer quotes.
  • For class ranking and top consumers, prefer mat_parse_report with org.eclipse.mat.api:overview and parse Class_Histogram*.txt.
  • For object inspection, use simple field-level OQL patterns (INSTANCEOF, OBJECTS 0x...).

Environment

All environment variables are optional:

  • MAT_HOME
  • MAT_LAUNCHER
  • JAVA_PATH (default java)
  • MAT_XMX_MB (default 4096)
  • MAT_TIMEOUT_SEC (default 1800)
  • MAT_CONFIG_DIR (default /tmp/mat-config)
  • MAT_DATA_DIR (default /tmp/mat-workspace)
  • MAT_DEBUG (default false)
  • MAT_DEBUG_LOG_DIR (default /tmp/mcp-mat-logs)
  • MAT_PRIVACY_MODE (default false)
  • MAT_OQL_MAX_BYTES (default 16384)
  • MAT_RESULT_PREVIEW_LINES (default 20)
  • MAT_STDIO_TAIL_CHARS (default 4000)

Run

npm install
npm run build
MAT_HOME=/path/to/mat node dist/src/server.js

Test

npm test

Install via npx

No build step required. Install directly from npm:

Claude Code

Add to ~/.claude/settings.json or project .claude/settings.json:

{
  "mcpServers": {
    "mat": {
      "command": "npx",
      "args": ["-y", "mcp-mat"],
      "env": {
        "MAT_HOME": "/Applications/MemoryAnalyzer.app/Contents/Eclipse"
      }
    }
  }
}
Codex
codex mcp add \
  --env MAT_HOME=/Applications/MemoryAnalyzer.app/Contents/Eclipse \
  mat \
  npx -y mcp-mat

Install in Claude Code (from source)

Add the MCP server to Claude Code settings (~/.claude/settings.json or project .claude/settings.json):

{
  "mcpServers": {
    "mat": {
      "command": "node",
      "args": ["/path/to/mcp-mat/dist/src/server.js"],
      "env": {
        "MAT_HOME": "/Applications/MemoryAnalyzer.app/Contents/Eclipse"
      }
    }
  }
}

Install in Codex (from source)

Build first:

cd /path/to/mcp-mat
npm install
npm run build

Add MCP server to Codex:

codex mcp add \
  --env MAT_HOME=/Applications/MemoryAnalyzer.app/Contents/Eclipse \
  mat \
  node /path/to/mcp-mat/dist/src/server.js

Verify:

codex mcp list
codex mcp get mat --json

If you need to update the entry:

codex mcp remove mat
# then run codex mcp add ... again

Equivalent ~/.codex/config.toml entry:

[mcp_servers.mat]
command = "node"
args = ["/path/to/mcp-mat/dist/src/server.js"]

[mcp_servers.mat.env]
MAT_HOME = "/Applications/MemoryAnalyzer.app/Contents/Eclipse"

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.