GitHub hosts HunyuanVideo, Tencent's open-source framework for large-scale video generation models, enabling AI-driven video creation.
MCPulse
$49/moStarting price
0Popularity
About MCPulse
MCPulse provides analytics for MCP servers by tracking how AI agents interact with tools. It records first-call success rates, retry behavior, empty responses, and schema costs without accessing arguments or results. The tool highlights tools that agents struggle with, those returning empty results, and unused tools that still consume schema tokens. Metrics include call volumes, latency, response sizes, and cost per session across different clients like Claude Desktop, Cursor, and Claude Code. The dashboard surfaces tools needing attention through colored rows and offers 16 metrics per tool. MCPulse compares tool descriptions against 4,749 public servers to identify inconsistencies in how tools are interpreted by agents. It does not log session content or sensitive data, only timing, size, and a hashed fingerprint of arguments.
Key features
- First-call success rate tracking
- Retry behavior analysis
- Empty response detection
- Schema cost measurement
- Latency and response size monitoring
- Tool health scoring against public servers
- Client-specific metrics (Claude Desktop, Cursor, etc.)
- 16 metrics recorded per tool
Use cases
- Debugging MCP tools that agents fail to use correctly
- Identifying unused tools consuming schema tokens
- Optimizing tool selection based on agent behavior
Pros
- Tracks first-call success and retry rates for MCP tools
- Identifies tools returning empty results or never called
- Measures schema costs and latency per tool
- Compares tool descriptions against public server ecosystem
- No access to arguments, results, or session content
Cons
- Limited to 10,000 calls per month on free plan
- No free tier for multiple servers
- Individual call logs retained for 7 days only on free plan
- Requires installation within MCP server environment
Frequently asked questions about MCPulse
What does MCPulse do?
MCPulse provides analytics for MCP servers by tracking how AI agents interact with tools, focusing on first-call success rates, retry behavior, empty responses, and schema costs without accessing arguments or results.
Who is MCPulse designed for?
MCPulse is designed for developers and teams managing MCP servers who need to understand how AI agents interpret and use their tools, identifying issues like poor first-call success or unused tools consuming schema tokens.
Does MCPulse log sensitive data?
No, MCPulse does not log session content or sensitive data. It only records timing, response size, and a hashed fingerprint of arguments to ensure privacy.
How does MCPulse compare to other MCP analytics tools?
MCPulse focuses on tool health metrics like first-call success and retry rates, and compares tool descriptions against 4,749 public servers, unlike tools that log arguments or session replays.
Can MCPulse identify tools that are never called?
Yes, MCPulse highlights tools that are registered but never used, which still consume schema tokens in every session, helping teams optimize their MCP server configurations.
How do I get started with MCPulse?
To get started, install MCPulse by adding two lines to your MCP server and sign up for a free account to access the dashboard and analytics.
