# Memento-Teams/Memento

Official Code of Memento: Fine-tuning LLM Agents without Fine-tuning LLMs

Repository: https://github.com/Memento-Teams/Memento
Canonical: https://ross.abutalabs.com/products/memento
Homepage: https://arxiv.org/abs/2508.16153
Language: Python
License: MIT
License Family: permissive
Last push: 2025-10-05T14:34:18+00:00

## Health v2 (maintenance only)
Score: 39/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 45, release rhythm 35, longevity 31
- inputs: {"age_days": 439, "days_push": 332, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2568, forks 297 (observed 2026-08-28T04:07:01.239882+00:00)

## What it is
Memento is a Python framework for building LLM agents that continually improve from experience via memory-based case-based reasoning, without fine-tuning model weights. It uses a planner-executor architecture with MCP tooling and episodic memory, achieving top results on the GAIA deep research benchmark.

## Use cases
- build llm agents that learn from past tasks without fine-tuning
- deep research agent that answers complex multi-step questions
- memory-based continual learning for ai agents
- agent that retrieves relevant past cases to solve new tasks
- orchestrate mcp tools with a planner-executor agent
- improve agent accuracy on out-of-distribution tasks via episodic memory

## When to choose
- you want adaptive agents without gradient updates or model weight changes
- you need a deep research agent with state-of-the-art GAIA performance
- you want case-based reasoning and memory augmentation in an agent loop
- you prefer MCP-based tool orchestration in Python

## When to avoid
- you need a simple single-shot chatbot without memory or learning
- you require fine-tuning of model weights rather than memory-based adaptation
- you need a production enterprise agent platform with vendor support

## Facets
- artifact type: framework
- maturity: active
- function: agent-framework, rag, llm-inference, mcp, machine-learning
- domain: artificial-intelligence, large-language-models, deep-learning
- platform: python, self-hosted
- tags: case-based-reasoning, memory-augmented-learning, continual-learning, planner-executor, deep-research, gaia-benchmark, online-reinforcement-learning, ai-agents, retrieval-augmented-generation, docker

## Member repositories
- Memento-Teams/Memento (main) score 39

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:01.239882+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:23:21.201237+00:00, confidence not recorded.
  - readme: https://github.com/Memento-Teams/Memento (fetched 2026-08-28T04:07:01.239882+00:00, sha 3632b8f15775)
  - homepage: https://arxiv.org/abs/2508.16153 (fetched 2026-08-29T10:05:46.887810+00:00, sha 6143751e506a)
  - site_page: https://info.arxiv.org/about/donate.html (fetched 2026-08-29T10:05:46.890545+00:00, sha cca9c3a11c56)
  - site_page: https://info.arxiv.org/about/ourmembers.html (fetched 2026-08-29T10:05:46.894051+00:00, sha 47cbc55ff1de)
  - site_page: https://info.arxiv.org/about (fetched 2026-08-29T10:05:46.896378+00:00, sha a1f16f915a9a)
  - site_page: https://info.arxiv.org/labs/index.html (fetched 2026-08-29T10:05:46.892376+00:00, sha b14a8d05a0ec)
- Data as of 2026-08-30T08:39:29.467469+00:00.