memoralabs 1.0.1

Python SDK for MemoraLabs — persistent memory for AI agents

memoralabs

Python SDK for MemoraLabs — persistent, self-improving memory for AI agents.

Store and recall memories in 5 lines of Python. Memory retrieval improves over time using Q-learning and temporal decay, so your agents get smarter, not just bigger.

Installation

pip install memoralabs

Requires Python 3.9+ and installs only one dependency: httpx.

Quick start

from memoralabs import MemoraLabs
client = MemoraLabs(api_key="ml_your_key")
memory = client.add("User prefers dark mode", user_id="user_42")
results = client.search("display preferences", user_id="user_42")
print(results.results[0].text)

Get an API key at memoralabs-api.onrender.com/docs.

All methods

Method	Description	Returns
signup(name, email, plan)	Register a new developer account	SignupResponse
rotate_key()	Rotate your API key (revokes old key)	KeyRotateResponse
add(text, *, user_id, agent_id, session_id, metadata)	Store a new memory	MemoryResponse
get(memory_id)	Retrieve a memory by ID	MemoryResponse
list(*, page, per_page, user_id, agent_id, session_id)	Paginated list of memories	MemoryListResponse
update(memory_id, *, text, metadata)	Update a memory's text or metadata	MemoryResponse
delete(memory_id)	Delete a memory	bool
search(query, *, user_id, agent_id, session_id, limit, min_score)	Semantic search	SearchResponse
gaps()	Detect knowledge gaps in stored memories	GapResponse
billing_status()	Current plan, usage counts, billing dates	BillingStatusResponse
checkout(*, success_url, cancel_url)	Create a Stripe Checkout session for Pro	CheckoutResponse

All optional keyword arguments default to None and are omitted from requests when not provided.

Error handling

from memoralabs import MemoraLabs, AuthError, PlanLimitError, NotFoundError

client = MemoraLabs(api_key="ml_your_key")

try:
    results = client.search("user preferences", user_id="user_42")
except AuthError:
    print("Invalid API key — check your credentials")
except PlanLimitError as e:
    print(f"Plan limit reached. Upgrade at: {e.upgrade_url}")
except NotFoundError:
    print("Memory not found")

All SDK exceptions inherit from MemoraLabsError and expose:

• status_code — HTTP status returned by the API
• error_code — Machine-readable error string (e.g. "PLAN_LIMIT_EXCEEDED")
• message — Human-readable description

Type errors are raised before any HTTP request is made, so you catch mistakes at call time rather than from network responses.

Context manager

with MemoraLabs(api_key="ml_your_key") as client:
    client.add("Session started", user_id="user_42")
# Connection pool released automatically

Namespacing memories

Use user_id, agent_id, and session_id to isolate memories across users, agents, and conversations:

# Store memory for a specific user and agent
client.add(
    "Prefers Python over JavaScript",
    user_id="user_42",
    agent_id="coding-assistant",
    session_id="session_101",
)

# Search only within that user's memories
results = client.search("language preference", user_id="user_42")

Full API reference

memoralabs-api.onrender.com/docs