> ## Documentation Index
> Fetch the complete documentation index at: https://memproof.kyberon.dev/llms.txt
> Use this file to discover all available pages before exploring further.
# Introduction
> Framework-agnostic memory control layer for AI agents
# What is Memproof?
Memproof is a native library (Python & TypeScript) that sits between your AI agent and its memory backend. Every memory operation — create, update, delete, search — passes through a control path that enforces policy, assesses risk, and produces an auditable event trail.
Memproof does **not** replace your memory store. It wraps it with governance.
## The Problem
Agent memory is now common across frameworks, but production teams face gaps:
* **Unsafe writes/deletes** are not consistently controlled
* **Memory APIs differ** by framework and provider
* **Audit evidence is fragmented** and hard to reconstruct during incidents
* **Policy enforcement** and human approval are rarely coupled to memory mutations
## How Memproof Solves This
```python theme={"theme":{"light":"github-light","dark":"github-dark-dimmed"}}
from memproof import Memproof
mp = Memproof(policy="./memproof.yaml", adapter="langgraph")
# Every mutation goes through the control path
result = await mp.remember(
content="user prefers dark mode",
scope={"tenant_id": "acme", "project_id": "chatbot", "agent_id": "agent-1"},
context={"actor_type": "agent", "actor_id": "agent-1",
"source": "langgraph", "timestamp": "2026-01-01T00:00:00Z"},
)
# result.status: committed | blocked | pending_approval | quarantined
# result.decision: allow | deny | require_approval | quarantine
# result.risk_assessment: score, level, factors
```
Every call produces:
1. A **risk assessment** (PII detection, secret detection, source trust, scope anomalies)
2. A **policy decision** (evaluated against your YAML rules)
3. An **immutable audit trail** (every stage of the pipeline logged via Trailproof)
## Key Properties
Works with LangGraph, OpenAI Sessions, MCP memory servers, or any custom backend via the adapter interface.
Native libraries for Python and TypeScript. Import and call in-process. No HTTP server required.
YAML-based rules with priority ordering, 10 condition operators, and deterministic decisions.
Every operation emits immutable lifecycle events to a [Trailproof](https://trailproof.kyberon.dev/docs/introduction) audit trail with SHA-256 hash chains and optional HMAC signing.
## Architecture at a Glance
## Next Steps
Install Memproof and run your first policy-controlled memory operation in 5 minutes.
See current OSS beta readiness status, quality gates, and performance baselines.