Quickstart
Store your first memory in under 5 minutes.
Prerequisites
- Python 3.9+
- A Mnemora API key (get one from the dashboard)
Install the SDK
pip install mnemora
Store your first memory
from mnemora import MnemoraSync
with MnemoraSync(api_key="mnm_...") as client:
client.store_memory("agent-1", "The user prefers concise replies.")
Search memories
from mnemora import MnemoraSync
with MnemoraSync(api_key="mnm_...") as client:
results = client.search_memory("user preferences", agent_id="agent-1")
for r in results:
print(r.content, r.similarity_score)
Full working example
This script stores agent state, writes a semantic memory, logs a conversation episode, then searches across all memories.
from mnemora import MnemoraSync
with MnemoraSync(api_key="mnm_...") as client:
# Working memory — fast key-value state
client.store_state("agent-1", {"task": "summarize quarterly report"})
# Semantic memory — auto-embedded, vector-searchable
client.store_memory("agent-1", "The user prefers bullet points over paragraphs.")
# Episodic memory — time-stamped event log
client.store_episode(
agent_id="agent-1",
session_id="sess-001",
type="conversation",
content={"role": "user", "message": "Summarize the Q3 results."},
)
# Vector search
results = client.search_memory("output format preferences", agent_id="agent-1")
for r in results:
print(f"[{r.similarity_score:.2f}] {r.content}")
# Retrieve state
state = client.get_state("agent-1")
print(state.data)
Expected output:
[0.94] The user prefers bullet points over paragraphs.
{'task': 'summarize quarterly report'}
Environment variable
Set MNEMORA_API_URL to override the default API endpoint — useful for self-hosted or staging deployments.
export MNEMORA_API_URL=https://your-custom-endpoint.example.com
Next steps
- Core concepts — understand the four memory types
- API reference — all 19 endpoints
- LangGraph integration — persistent graph checkpoints
- LangChain integration — chat message history
- CrewAI integration — agent storage backend