AI Agent Observability: Track Every Tool Call with ReskPoints

#observability #ai #opensource #python

Links

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

AI agents are hard to debug. When your agent makes a tool call, calls an API, or reads a file, where does that show up? Most teams rely on print statements or build custom logging for each integration — and that does not scale.

ReskPoints changes that. It is an open-source AI Agent Logger that gives you structured, exportable traces of every action your agent takes.

Quick Start

pip install reskpoints

Create a logger:

from reskpoints import AgentLogger

logger = AgentLogger(
    service_name="my-agent",
    sampling_rate=1.0,  # log every action
    export="console"
)

# Log any action your agent performs
logger.log_action(
    action="tool_call",
    tool="web_search",
    parameters={"query": "latest AI news"},
    result="3 results returned",
    duration_ms=450
)

Multi-Export in One Line

Switch to Prometheus or Datadog without changing your code:

from reskpoints import AgentLogger

logger = AgentLogger(
    service_name="production-agent",
    export="prometheus",
    prometheus_port=8000
)

Or send to multiple backends:

logger = AgentLogger(
    service_name="multi-export-agent",
    export="datadog+console+opentelemetry"
)

Smart Sampling

ReskPoints supports configurable sampling so you control the volume — log every Nth call, or every call above a configurable threshold. Masking redacts sensitive fields automatically before they hit your observability pipeline.

Why This Matters

70 percent of organisations lack AI governance according to PwC. Visibility into agent behaviour is the first step. ReskPoints gives you that visibility in five minutes.