mnemo-sdk 0.1.0

Memory and observability for AI agents. Two lines of code to give any agent persistent memory and full trace visibility.

mnemo

Memory and observability for AI agents. Two lines of code.

from mnemo import MnemoClient

mnemo = MnemoClient(tenant_id="your-tenant")
result = mnemo.run(agent_id="my-agent", prompt="Hello, world!")

That's it. Your agent now has persistent memory and full trace visibility.

Install

pip install mnemo

# With all integrations
pip install mnemo[all]

# Pick what you need
pip install mnemo[anthropic,mem0,langfuse]

What happens when you call mnemo.run()

1. Retrieves relevant memories from past runs
2. Injects them into the prompt context
3. Calls your LLM (Claude, OpenAI, or any custom function)
4. Logs the full trace to Langfuse + OpenTelemetry
5. Extracts new memories from the response
6. Checks alert rules and fires if needed
7. Returns the response — unchanged

Configuration

from mnemo import MnemoClient
from mnemo.client import LLMConfig

mnemo = MnemoClient(
    tenant_id="acme-corp",
    llm=LLMConfig(provider="anthropic", model="claude-sonnet-4-20250514"),
    debug=True,
)

Or use environment variables:

export MNEMO_MEM0_API_KEY=your-key
export MNEMO_LANGFUSE_PUBLIC_KEY=your-key
export MNEMO_LANGFUSE_SECRET_KEY=your-key
export ANTHROPIC_API_KEY=your-key

License

MIT