Track and analyze LLM API costs, token consumption, and performance metrics in real-time for your AI agent.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agentpulse
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install agentpulse using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is AgentPulse?
AgentPulse is a specialized monitoring and analytics skill designed for developers who need deep visibility into their AI agent's operational efficiency and expenses. By integrating this skill, you can gain immediate insights into spending across 50+ models from providers like Anthropic, OpenAI, and Google. It provides a centralized dashboard to visualize latency trends, error rates, and daily spending, ensuring your agent remains cost-effective and reliable.
As one of the essential Openclaw Skills for production environments, AgentPulse prioritizes security by sending only metadata, such as model names and token counts, to its dashboard. This allows you to manage budgets and optimize model selection without exposing sensitive prompt content or conversation data. Whether you are scaling a startup or managing personal projects, this skill bridges the gap between raw API calls and actionable financial intelligence.
AgentPulse Use Cases
- Monitoring daily spending limits to prevent unexpected LLM API bills.
- Analyzing model performance by comparing average latency across different providers.
- Debugging agent failures by tracking error rates and specific API status codes.
- Identifying the most cost-effective models for specific agent tasks.
- Generating reports on token usage for internal audits or client billing.
How AgentPulse Works
- The agent executes an LLM API call using its configured provider (e.g., Anthropic or OpenAI).
- Post-execution, the skill captures metadata including input/output tokens, latency in milliseconds, and the success status.
- This data is reported silently to the AgentPulse backend via a secure HTTPS POST request.
- When a user queries about performance or costs, the skill fetches aggregated data from the stats overview endpoint.
- The agent presents the summarized financial and technical data to the user in a clear, conversational format.
AgentPulse Setup
First, sign up for a free account at agentpulse.dev and obtain your API key from the settings page. Set the environment variable on the machine running your agent:
export AGENTPULSE_API_KEY="ap_your_api_key_here"
Alternatively, you can add it to your configuration file to ensure it persists across sessions within your Openclaw Skills environment:
# Edit ~/.openclaw/openclaw.json
# Add to: skills.entries.agentpulse.env.AGENTPULSE_API_KEY
Finally, restart your session to activate the skill with the new credentials.
AgentPulse Data Schema & Taxonomy
AgentPulse organizes data into event-based records and aggregated statistics. All costs are calculated server-side using up-to-date pricing tables.
| Field | Description |
|---|---|
| agent_name | The unique identifier for the reporting agent |
| provider | The LLM provider (e.g., anthropic, openai, deepseek) |
| model | The exact model string used for the API call |
| input_tokens | Total tokens sent in the prompt |
| output_tokens | Total tokens received in the completion |
| latency_ms | Round-trip time for the API call in milliseconds |
| status | The result of the call (success or error) |
| error_message | The specific error string if the call failed |
AgentPulse Advanced Features
- Automated server-side cost calculation for over 50 models including Claude 3.5 Sonnet and GPT-4o.
- Configurable alerts for daily cost thresholds, consecutive failures, and rate limit spikes.
- Multi-model tracking support with real-time pricing updates for DeepSeek, Gemini, and Mistral.
- Comprehensive API for fetching daily breakdowns, top models by spend, and average latency stats.
- Background reporting mode that logs events without interrupting the user conversation flow.
SKILL.md
Loading
Related Openclaw Skills
Solana Sniper Bot
An autonomous Solana trading bot that monitors liquidity pools and uses AI to detect rugpulls before executing trades.
Binance Spot Trader
An autonomous trading bot that combines technical indicators with LLM-powered market sentiment analysis to execute trades on the Binance spot market.
Clawflow Multi-Agent Orchestration Protocol
Clawflow is a recursive task delegation protocol that enables Openclaw Skills to collaborate through message-passing and Directed Acyclic Graphs (DAGs).
Medical Revenue Audit
An automated auditing tool that cross-references surgical logs with billing records to identify and recover lost revenue.
QR Auction Bidder
Automate USDC bids on daily $QR code auctions on the Base network using the Bankr transaction layer.
System Diagnostics
A comprehensive system diagnostics and troubleshooting toolkit designed for DevOps engineers and system administrators using Openclaw Skills.
Featured*
