Track Every AI Agent Action with ReskPoints -- Agent Logger for Python

Links:

• GitHub: https://github.com/Resk-Security/ReskPoints
• PyPI: https://pypi.org/project/reskpoints
• resk.fr: https://resk.fr

The Problem

When your AI agent calls a tool, how do you know what happened? Print statements get lost. Logging frameworks are too heavy for agent loops. You need something purpose-built.

Enter ReskPoints

ReskPoints is a lightweight Python library that logs every agent action with sampling, masking, and multi-export support.

Quick Start

pip install reskpoints

from reskpoints import AgentLogger

logger = AgentLogger(
    sample_rate=0.5,       # log 50 percent of actions
    mask_fields=['api_key', 'password'],
    exporters=['console', 'prometheus']
)

@logger.track
def call_llm(prompt: str):
    # your agent logic here
    return response

call_llm('analyze this report')
# -> logged with duration, tokens, success/failure

Key Features

• Sampling -- log every Nth action or a percentage. No firehose.
• Masking -- strip sensitive fields before they hit the log.
• Multi-export -- Datadog, Prometheus, OpenTelemetry, console, or file. No code changes to switch.
• Token accounting -- track token spend per agent step.
• Production-ready -- built for loops, not debugging sessions.

When to Use It

• Monitoring multi-step agent pipelines
• Debugging tool call failures in production
• Billing or usage tracking per user
• Compliance logging for AI systems

Try it out: pip install reskpoints and check the docs at resk.fr.

ReskPoints is Apache 2.0 licensed. Contributions welcome on GitHub.