1. Conduid
  2. Other
  3. ClaudePulse
MCP server · Other

ClaudePulse

Unclaimed last commit 6 months ago other
30Low

Scored 6 months ago · breakdown

About ClaudePulse

ClaudePulse is an MCP server published by ccivlcid in the Other category.

The repository has 3 stars and 0 forks, with the last commit 6 months ago. Six months or more without a commit doesn't mean the server is broken, but check the open issues before depending on it in production.

Install

From source
git clone https://github.com/ccivlcid/ClaudePulse

This server has no ConduID identity, so agent calls to it are not receipted. Pin the version you install and review the source before granting it credentials.

Tools (13)

pulse_agent_statusActive agent status

Parameters not published.

pulse_file_heatmapFile access frequency ranking

Parameters not published.

pulse_open_dashboardOpen the web dashboard

Parameters not published.

pulse_reset_allReset all Pulse data

Parameters not published.

pulse_reset_sessionDelete a specific session

Parameters not published.

pulse_server_errorsDev server errors only

Parameters not published.

pulse_server_healthServer health diagnosis

Parameters not published.

pulse_server_logsRecent dev server logs

Parameters not published.

pulse_session_statsCurrent session statistics

Parameters not published.

pulse_start_serverStart dev server + log capture

Parameters not published.

Show all 13 tools
pulse_stop_serverStop dev server

Parameters not published.

pulse_timelineSession timeline summary

Parameters not published.

pulse_token_usageEstimated token usage

Parameters not published.

Ask AI

✦

Ask AI about ClaudePulse

Powered by Claude · Grounded in docs

I know everything about ClaudePulse. Ask me about installation, configuration, usage, or troubleshooting.

Security checks

  • ✓README presentThe repository ships documentation.

5 checks have not been run yet.

README


At a Glance

Every file read, edit, and test that Claude Code performs is automatically recorded and viewable in real time — either from the CLI or a web dashboard.

Claude Code CLI
  │
  ├── Hooks auto-record every tool use  →  ~/.claude-pulse/sessions/*.jsonl
  │
  ├── Query from CLI          "/claude-pulse:stats"
  │
  └── Visualize in dashboard  "/claude-pulse:dashboard"  →  http://localhost:52101
  │
  └── Start/stop dev server   "/claude-pulse:server <command>"  →  http://localhost:<port>

Installation

# 1. Register the marketplace (one-time)
claude plugin marketplace add ccivlcid/ClaudePulse

# 2. Install the plugin
claude plugin install claude-pulse@claude-pulse

# 3. Enable the plugin
claude plugin enable claude-pulse

Data collection begins immediately after installation. No additional configuration required.

Requirements

  • Node.js >= 20
  • Claude Code CLI

Plugin Management

# Uninstall
claude plugin uninstall claude-pulse

# Uninstall (preserve data — keeps ~/.claude-pulse/)
claude plugin uninstall claude-pulse --keep-data

# Update
claude plugin update claude-pulse

# Disable / re-enable
claude plugin disable claude-pulse
claude plugin enable claude-pulse

Use scopes to control where the plugin is active:

Scope Flag Applies to
user --scope user (default) All projects
project --scope project Current project only (.claude/settings.json)
local --scope local Local only (.claude/settings.local.json, untracked)

Usage

Slash Commands

Command Description
/claude-pulse:stats Current session statistics summary
/claude-pulse:dashboard Open the web dashboard
/claude-pulse:logs View dev server logs
/claude-pulse:server Start / stop a dev server
/claude-pulse:errors All session errors (tool failures + server errors)
/claude-pulse:heatmap File access frequency ranking
/claude-pulse:agents Active agent status
/claude-pulse:tokens Estimated token usage
/claude-pulse:timeline Session timeline summary
/claude-pulse:health Dev server health diagnosis
/claude-pulse:reset Reset session data or all data

1. Session Statistics

> /claude-pulse:stats

── Session Stats ──────────────────
Elapsed: 6m | Tools: 67 | Errors: 1

  Read: 32  Edit: 12  Bash: 8
  Grep: 8   Agent: 4  Write: 3
───────────────────────────────────

Natural language also works:

  • "Show me the session stats"
  • "How much have I used so far?"
  • "Which files were modified the most?"

2. Web Dashboard

> /claude-pulse:dashboard

Opens http://localhost:52101 in your browser automatically.

Dashboard panels:

Panel Description
Activity Stream Real-time chronological feed of all tool usage
Alert Center Combined view of tool errors and server errors
Agent Tracker Main agent and subagent activity tracking
Top Files Most frequently accessed files, ranked
Token Usage Estimated token consumption per session (tool-based)
Server Monitor Live dev server log output

Click the ↗ button in any panel header to pop it out into a full-screen window.

The dashboard uses SSE (Server-Sent Events) for real-time updates. Claude Code activity is reflected instantly without refreshing.

3. Dev Server Monitoring

Start or stop a dev server using slash commands or natural language.

> /claude-pulse:server start npm run dev
→ "Dev server started (PID: 12345, localhost:3000)"

> /claude-pulse:server stop
→ "Dev server stopped"

Natural language works too:

> Start the dev server
Claude Code → pulse_start_server({ command: "npm run dev" })
→ "Next.js server started (localhost:3000)"

> Stop the dev server
Claude Code → pulse_stop_server()
→ "Server stopped"

While the server is running, logs are automatically captured and displayed in the Server Monitor panel in real time.

Checking errors:

> Any server errors?
→ "1 found — TypeError at src/api/auth.ts:42 (14:35:05)"
→ Claude Code automatically reads the file → suggests a fix

> /claude-pulse:errors
→ Summary of all errors in the session (tool failures + server errors)

Auto-detected error patterns:

  • JavaScript exceptions: TypeError, ReferenceError, SyntaxError
  • HTTP errors: 4xx/5xx responses
  • Build errors: Failed to compile, Module not found
  • Network errors: EADDRINUSE, ECONNREFUSED
  • Stack traces

How It Works

Data Collection (Automatic)

Claude Pulse uses the built-in Hook system in Claude Code.

User: "Refactor the auth module"

  Claude Code calls the Read tool
  → PreToolUse Hook fires
  → collect-event.mjs runs
  → One line appended to ~/.claude-pulse/sessions/{sessionId}.jsonl

  Claude Code calls the Edit tool
  → Same process repeats

  Tool failure
  → PostToolUseFailure Hook fires
  → Error details recorded

Collected events:

Hook Event Description
SessionStart session-start Session started
PreToolUse tool-start Tool use started
PostToolUse tool-end Tool use completed
PostToolUseFailure tool-error Tool execution failed
SubagentStart agent-start Subagent started
SubagentStop agent-stop Subagent stopped
SessionEnd session-end Session ended

Data Storage

All data is stored in local JSONL files. No external server or database required.

~/.claude-pulse/
├── sessions/
│   ├── {sessionId-1}.jsonl    # Per-session event logs
│   └── {sessionId-2}.jsonl
├── servers/
│   └── {sessionId}.jsonl      # Dev server logs
├── index.json                 # Session index (for fast lookups)
└── config.json                # Configuration

Example JSONL line:

{"id":"a1b2","ts":"2026-03-30T14:32:01Z","type":"tool-start","sessionId":"8ebad784-...","toolName":"Read","toolUseId":"toolu_01Gt...","filePath":"src/auth.ts","projectDir":"/home/user/my-app"}

Data Retention

  • Event logs: 30-day retention (default)
  • Server logs: 7-day retention
  • Max storage: 500 MB (oldest sessions deleted when exceeded)
  • Cleanup runs only at session end (zero overhead during work)

MCP Tools

These tools can be invoked via natural language inside Claude Code.

Tool Description Example
pulse_session_stats Current session statistics "Show me the session stats"
pulse_server_logs Recent dev server logs "Show the last 20 server logs"
pulse_server_errors Dev server errors only "Any server errors?"
pulse_file_heatmap File access frequency ranking "Which file was modified most?"
pulse_agent_status Active agent status "What's the agent status?"
pulse_token_usage Estimated token usage "How many tokens have I used?"
pulse_server_health Server health diagnosis "How's the server doing?"
pulse_start_server Start dev server + log capture "Start the dev server"
pulse_stop_server Stop dev server "Stop the dev server"
pulse_timeline Session timeline summary "Summarize today's work"
pulse_open_dashboard Open the web dashboard "Open the dashboard"
pulse_reset_session Delete a specific session "Reset session data"
pulse_reset_all Reset all Pulse data "Clear all Pulse data"

Configuration

Edit ~/.claude-pulse/config.json:

{
  "retention": {
    "eventsDays": 30,
    "serverLogsDays": 7,
    "maxTotalSizeMb": 500
  },
  "ports": {
    "dashboard": 52101
  }
}

You can also override the port via environment variable:

PULSE_DASHBOARD_PORT=52102

About Token Estimation

Token estimates are based on tool input text length (~4 chars/token) and per-tool average output token tables.

Since hooks cannot observe system prompts, conversation history, or reasoning tokens, only tool-related tokens can be estimated. The estimates are useful for relative comparisons across sessions and projects; absolute numbers are approximate.


Tech Stack

Component Technology
Event collection JSONL (fs.appendFileSync, zero dependency)
MCP Server @modelcontextprotocol/sdk
Dashboard API Hono
Real-time SSE (Server-Sent Events)
Frontend React 19 + Vite
Styling Tailwind CSS 4
Charts Recharts
State Zustand

No native C++ module dependencies — works on Windows, macOS, and Linux.


Limitations

  • Dev servers started directly from a terminal (outside Claude Code) cannot have their logs captured — use pulse_start_server instead
  • Token estimates are tool-call-based approximations (system prompt and conversation history excluded)
  • Hooks only report tool start/end — intermediate progress is not observable

License

MIT

README copy captured when this server was indexed; it may lag the repository. The original is authoritative.

Questions

About ClaudePulse

How do I install ClaudePulse?

No package has been published for this server, so install it from source: clone the repository and follow its README.

Is ClaudePulse safe to use with an AI agent?

Its trust score is 30 out of 100 (low). It passes 1 of the 1 static security checks run so far. It has no ConduID identity yet, so agent calls to it are not receipted.

Is ClaudePulse still maintained?

The last commit was 6 months ago. That's long enough that you should check whether the maintainer is responding to issues before depending on it.