AgentOps + MemorySync

Overview

opentelemetry-instrumentation-memorysync makes every MemorySync operation — adds, recalls, queries, forgets, and ten more — a first-class span in AgentOps traces. agentops.init() registers AgentOps as the global OpenTelemetry tracer provider, so one instrument_memorysync() call is the whole integration; the standard BaseInstrumentor entry point also works with the opentelemetry-instrument CLI for zero-code setups. Spans carry operation, sync/async client kind, result counts, score aggregates, idempotency signals, and error classifications — and, unlike the incumbent’s instrumentation, memory CONTENT is never recorded unless you opt in (and then truncated): the in-tree Mem0 wrapper dumps full memory text into span attributes and — via leftover debug print() calls — onto stdout on every add and search. Full sync/async parity across both MemorySync clients (the Mem0 wrapper skips history on cloud clients), guaranteed pass-through (results and exceptions are untouched; a broken attribute extractor can never fail your call), and a seam test pins the AgentOps global-provider contract on every push. Zep, Supermemory, and Letta have no AgentOps story at all.

Setup status and requirements

Supported version

opentelemetry-instrumentation-memorysync 1.0.0 (PyPI)

Last setup review

2026-08-25

Permissions

A MemorySync API key; an AgentOps API key for the AgentOps dashboard.

Limits

Python 3.9+. Works with any OpenTelemetry backend — AgentOps is one init() away.

Capabilities

• The FIRST SDK-level tracing instrumentor from any memory vendor
• One line: instrument_memorysync() after agentops.init() — order independent
• Standard OTel BaseInstrumentor + entry point — works with opentelemetry-instrument
• Privacy-first: memory content is NEVER recorded unless explicitly opted in
• Zero stdout pollution — the incumbent’s wrapper print()s memory content
• Full sync/async parity across 14 memory operations
• Instrumentation can never break the app — guaranteed pass-through
• Duration histogram per operation for latency dashboards

Quick Start

import agentops
from opentelemetry.instrumentation.memorysync import instrument_memorysync
from memorysync import MemorySyncClient

agentops.init()
instrument_memorysync()

client = MemorySyncClient(api_key="ms_...", base_url="https://api.memorysync.io")
client.add("Prefers window seats", source="chat")   # span: memorysync.add
client.query("seating preferences", k=5)            # span: memorysync.query

Use Cases

Explore More