Mnemoverse Python SDK

Persistent memory for AI agents. Not vector search — statistical learning backed by Hebbian associations.

Installation

pip install mnemoverse

Quick Start

export MNEMOVERSE_API_KEY=mk_live_YOUR_KEY

Check the key in one command. It reads the same variable the client reads (in PowerShell, write $env:MNEMOVERSE_API_KEY and type curl.exe):

curl -s -H "X-Api-Key: $MNEMOVERSE_API_KEY" https://core.mnemoverse.com/api/v1/memory/stats

from mnemoverse import MnemoClient

client = MnemoClient()
# or pass it explicitly — an explicit api_key always wins over the environment:
# client = MnemoClient(api_key="mk_live_YOUR_KEY")

# Store a memory
result = client.write(
    "Retry with exponential backoff fixed the timeout issue",
    concepts=["retry", "backoff", "timeout"]
)

# Query — Hebbian associations expand "timeout" → "retry", "backoff"
memories = client.read("how to handle timeouts?")

# Report outcome — the system learns what works
client.feedback(
    atom_ids=[item.atom_id for item in memories.items],
    outcome=1.0,
    query_concepts=memories.query_concepts
)

Closing the client

MnemoClient holds a background event loop and a pooled HTTP connection, so close it when a script is done. Use it as a context manager, with MnemoClient() as client:, to close automatically on exit, or call client.close() directly. A client left unclosed closes itself at interpreter exit, which is a fallback rather than something to rely on.

Async Client

from mnemoverse import AsyncMnemoClient

async with AsyncMnemoClient(api_key="mk_live_YOUR_KEY") as client:
    result = await client.write("async memory", concepts=["async"])
    memories = await client.read("what about async?")

Features

• Circuit breaker — 5 failures → open → 30s half-open → probe
• Retry with backoff — 3 attempts, rate-limit-aware
• Sync + async — MnemoClient for scripts, AsyncMnemoClient for FastAPI
• Type-safe — Pydantic models, full type hints

Methods

Every method exists on both MnemoClient (sync) and AsyncMnemoClient (async).

Search or feed?

read() answers what do I know about X and ranks by relevance. recent() answers what happened lately and is complete within one scope by construction — nothing is skipped, which a semantic search cannot promise. Reach for recent() to resume after a break or to catch up on a shared room.

from mnemoverse import MnemoClient

client = MnemoClient(api_key="mk_live_...")

# Catch up on a shared room. Rooms are SEPARATE stores: pass the address as
# `domain`, or an unscoped feed will not cover them.
page = client.recent(domain="xroom:room_01ABC", since="2026-08-01T00:00:00Z", limit=20)
for item in page.items:
    print(item.created_at, item.content)

if page.next_cursor:
    page = client.recent(domain="xroom:room_01ABC", cursor=page.next_cursor)

Read items carry created_at and provenance (who wrote it, where from).

Documentation

• Getting Started
• API Reference
• Python SDK Docs

License

MIT