Evaluation playground

Test the memory agent, live

Pick a benchmark question, see how your model answers from Memanto's memory, and have a second LLM judge it against ground truth.

Context Question

Choose a ground-truth question from the benchmarks. This sets the Agent ID.

Inference LLM

The model generating the initial answer.

Keys are not stored locally.
Judge LLM

The model evaluating the accuracy of the answer.

Ready to evaluate
Review the context question and click Send to test the agent.
Like the scores?

Put the same memory behind your agents

The agent you just tested is the one you install. Star the repo and give your own tools the memory that scored these answers.

Star on GitHubRead the research
bash
pip install memanto