v8-cpu-profile-decoder-mcp

v8-cpu-profile-decoder-mcp 🐸⚡

An MCP server that decodes V8 CPU profiles into token-efficient bottleneck summaries for AI agents.

Your Node.js app is slow. You ran --cpu-prof. Now you have a 20MB .cpuprofile file — and your AI agent is completely blind to it.

---

🤔 The Problem

V8 CPU profiles are massive. A typical .cpuprofile from a production Node.js app is 5–50MB of raw JSON — millions of lines mapping memory addresses, tick counts, and microsecond execution sequences. It looks like this:

{
  "nodes": [
    { "id": 1482, "callFrame": { "functionName": "processRequest", "url": "file:///app/dist/server.js", "lineNumber": 847 }, "hitCount": 3241, "children": [1483, 1490] },
    ...
  ],
  "samples": [1482, 1483, 1482, 1490, 1482, ...],
  "timeDeltas": [120, 98, 115, 102, ...]
}

An AI agent attempting to read this file instantly collapses its context window and fails. Even if it could read it, it can't run the aggregation algorithms needed to compute inclusive/exclusive CPU times across the call tree.

So when you ask your agent:

• 🙈 "Which function is consuming the most CPU?"
• 🙈 "What's calling my slow database query?"
• 🙈 "Which TypeScript file is the bottleneck actually coming from?"

...it's guessing. It has no access to the profiling data.

v8-cpu-profile-decoder-mcp fixes that. It decodes the profile locally and hands the agent a 10-line semantic summary instead of a 50MB file.

---

🛠️ Tools

extract_hottest_functions

Parses the .cpuprofile and returns the top N functions ranked by exclusive CPU time (self time). Filters out V8 internals and Node.js built-ins — only user code.

{
  "profile_path": "/app/profiles/CPU.20260516.cpuprofile",
  "top_n": 5,
  "min_self_percent": 1.0
}

[
  {
    "rank": 1,
    "functionName": "hashPassword",
    "url": "file:///app/dist/auth/crypto.js",
    "lineNumber": 42,
    "selfTimeMs": 1842.5,
    "totalTimeMs": 1842.5,
    "selfPercent": 61.32,
    "totalPercent": 61.32,
    "hitCount": 3241
  },
  {
    "rank": 2,
    "functionName": "parseJsonBody",
    "url": "file:///app/dist/middleware/body.js",
    "lineNumber": 18,
    "selfTimeMs": 412.1,
    "totalTimeMs": 412.1,
    "selfPercent": 13.71,
    "totalPercent": 13.71,
    "hitCount": 724
  }
]

---

analyze_call_tree_path

Finds all callers of a specific function and shows how often each one invoked it. Accepts partial, case-insensitive function name matching.

{
  "profile_path": "/app/profiles/CPU.20260516.cpuprofile",
  "function_name": "hashPassword",
  "top_callers": 3
}

{
  "targetFunction": "hashPassword",
  "matchedNodes": 2,
  "totalSelfTimeMs": 1842.5,
  "totalPercent": 61.32,
  "callers": [
    {
      "functionName": "loginHandler",
      "url": "file:///app/dist/routes/auth.js",
      "lineNumber": 94,
      "callCount": 2180,
      "selfTimeMs": 240.1
    },
    {
      "functionName": "validateSession",
      "url": "file:///app/dist/middleware/auth.js",
      "lineNumber": 31,
      "callCount": 1061,
      "selfTimeMs": 116.8
    }
  ]
}

---

correlate_source_code

Maps compiled JS bottlenecks back to their original TypeScript source locations using .js.map files. Falls back gracefully to compiled JS locations if no source map is found.

{
  "profile_path": "/app/profiles/CPU.20260516.cpuprofile",
  "top_n": 5
}

{
  "resolved": [
    {
      "rank": 1,
      "generatedUrl": "file:///app/dist/auth/crypto.js",
      "generatedLine": 42,
      "source": {
        "originalFile": "src/auth/crypto.ts",
        "originalLine": 38,
        "originalColumn": 2,
        "originalFunction": "hashPassword"
      },
      "selfTimeMs": 1842.5,
      "selfPercent": 61.32
    }
  ],
  "sourcemapErrors": []
}

analyze_gc_pressure

Reports garbage collection overhead as a percentage of profiling duration, broken down by GC type. Flags when GC exceeds a configurable threshold and provides a targeted recommendation.

{
  "profile_path": "/app/profiles/CPU.cpuprofile",
  "threshold_percent": 10
}

{
  "gc_ticks": 184,
  "total_ticks": 1240,
  "gc_percentage": 14.84,
  "gc_type_breakdown": {
    "scavenger": 122,
    "mark_sweep": 0,
    "mark_compact": 0,
    "incremental": 62,
    "generic": 0
  },
  "exceeds_threshold": true,
  "threshold_percent": 10,
  "verdict": "GC consumed 14.84% of CPU — exceeds the 10% threshold. Dominated by Scavenger (short-lived object pressure). Consider object pooling, reusing buffers, or reducing closure captures."
}

---

diff_profiles

Compares two .cpuprofile files (before/after an optimization) and returns per-function CPU time deltas, normalized against each profile's total duration. Frames are matched by call-frame coordinates, not transient node IDs, so alignment is stable across profiling sessions.

{
  "before_profile_path": "/app/profiles/before.cpuprofile",
  "after_profile_path": "/app/profiles/after.cpuprofile",
  "top_n": 5
}

{
  "before_duration_ms": 5000,
  "after_duration_ms": 4800,
  "total_execution_delta_ms": -200,
  "total_execution_delta_percent": -4,
  "top_improvements": [
    {
      "function_name": "hashPassword",
      "url": "file:///app/dist/auth/crypto.js",
      "line_number": 42,
      "before_ms": 1842.5,
      "after_ms": 620.1,
      "absolute_diff_ms": -1222.4,
      "relative_diff_percent": -66.34
    }
  ],
  "top_regressions": [],
  "only_in_before": [],
  "only_in_after": []
}

---

analyze_async_bottlenecks

Detects event-loop overhead by identifying V8 internal frames representing async machinery — microtask queue processing, nextTick saturation, and timer/immediate callbacks.

{
  "profile_path": "/app/profiles/CPU.cpuprofile",
  "threshold_percent": 10
}

{
  "total_ticks": 1240,
  "async_ticks": 186,
  "event_loop_overhead_ms": 372,
  "event_loop_overhead_percent": 15.0,
  "dominant_async_patterns": [
    { "pattern": "promise_chains", "ticks": 142, "percent": 11.45 },
    { "pattern": "nexttick_saturation", "ticks": 44, "percent": 3.55 }
  ],
  "verdict": "Event-loop overhead is 15.0% of CPU — exceeds the 10% threshold. Promise chain overhead is visible in the profile. Consider batching microtasks, using Promise.all() to parallelise I/O, or offloading CPU-bound continuations to worker threads."
}

---

🚀 Installation

npx v8-cpu-profile-decoder-mcp

Or install globally:

npm install -g v8-cpu-profile-decoder-mcp

Generate a CPU profile in Node.js

# Single run
node --cpu-prof your-script.js

# With custom output dir
node --cpu-prof --cpu-prof-dir ./profiles your-script.js

Or programmatically via Chrome DevTools → Performance tab → Record.

Claude Desktop config

{
  "mcpServers": {
    "v8-cpu-profile-decoder-mcp": {
      "command": "npx",
      "args": ["-y", "v8-cpu-profile-decoder-mcp"]
    }
  }
}

---

💡 Example Agent Prompts

"Here's my CPU profile at /app/profiles/CPU.cpuprofile — which function is consuming the most CPU?"

"Find what's calling processRequest in this profile and how often"

"Map the top 10 hottest functions back to their original TypeScript files"

"My Node.js API is slow under load — profile is at /tmp/CPU.cpuprofile, find the bottleneck"

"Is GC the bottleneck? Check the profile at /tmp/CPU.cpuprofile and tell me what kind of allocation is causing it"

"Compare these two profiles before and after my optimization — which functions improved and which regressed?"

"Is this app spending too much CPU on async overhead and event-loop machinery?"

---

🔗 Related Projects

• playwright-trace-decoder-mcp — decode Playwright traces for CI failure root-cause analysis
• playwright-network-chaos-mcp — simulate network failures and latency in browser sessions
• flakiness-knowledge-graph-mcp — knowledge graph of flaky test patterns
• ast-impact-mapper-mcp — find affected tests from code changes via TypeScript AST
• playwright-spatial-layout-mcp — geometric spatial awareness of web layouts

---

📄 License

MIT © vola-trebla