AI use cases RAG, knowledge & memory Long-term memory for AI agents
Long-term memory for AI agents
Give AI agents and assistants memory that persists across sessions: preferences, facts and past decisions. Workflow, pitfalls and ranked memory tools.
- Tools compared
- 2
- Open source
- 2of 2
- Reviewed
- 16 Oct
For: Developers
- 1Mem0AI tool55Caution
- 2LlamaIndexAgent framework43Caution
What the score and shapes mean
Score = public maintenance, adoption and docs signals, as a percentile of 52 tools. Nobody here tested these hands-on. Shapes are the separate safety label: No issues found, No dependencies scanned, Caution.
What it does for you
Developers building assistants and agents that talk to the same people repeatedly.
- Remember user preferences between sessions
- Store and recall facts from past conversations
- Keep separate memory per user or per project
- Update or delete outdated memories
- Add personalisation to a support or coaching assistant
More about this job
Without memory, every chat starts from zero. A memory layer stores facts and preferences from conversations and retrieves the relevant ones later, so an assistant can stay consistent across sessions and users.
Workflow in 5 steps
- Decide what is worth remembering and what must never be stored.
- Add a memory layer and scope memories by user id.
- Retrieve a small number of relevant memories per request, not everything.
- Give users a way to view, edit and delete what is stored.
- Test for stale or wrong memories and set expiry rules.
Full steps
- Decide what is worth remembering and what must never be stored.
- Add a memory layer and scope memories by user id.
- Retrieve a small number of relevant memories per request, not everything.
- Give users a way to view, edit and delete what is stored.
- Test for stale or wrong memories and set expiry rules.
Prompts to copy
Starting points we wrote, not tested results.
From this conversation, extract durable facts about the user (preferences, constraints, goals). Output a list; skip anything temporary or sensitive.
2 more prompts
Given these stored memories and the new message, which memories are relevant? Return their ids and a one-line reason.
These two memories contradict each other. Decide which is newer and propose an update.
Watch out for
- Prompts sent to a hosted model leave your machine.
- AI output can be fluent and wrong.
- Stored memories are personal data in many jurisdictions.
- CautionMem018 critical-severity known vulnerabilities in runtime dependencies (OSV)
- CautionLlamaIndex39 critical-severity known vulnerabilities in runtime dependencies (OSV)
Why these warnings?
- Prompts sent to a hosted model leave your machine. For confidential material use a local model or check the provider's data terms first.
- AI output can be fluent and wrong. Check facts, code and citations yourself before you rely on them or publish them.
- Stored memories are personal data in many jurisdictions. Plan for access, correction and deletion requests from day one.
- Mem0: 18 critical-severity known vulnerabilities in runtime dependencies (OSV); Skill/agent file instructs piping a remote script into a shell: skills/mem0-oss-to-platform/references/gotchas.md — curl -fsSL https://raw.githubusercontent.com/mem0ai/mem0/main/scripts/oss-to-platform-migrate.sh | bash
- LlamaIndex: 39 critical-severity known vulnerabilities in runtime dependencies (OSV)
Recommended tools
From our ranked data, highest score first. A high score does not mean a clean scan.
- 55Mem0AI tool · Apache-2.0 · #21 overall
- 43LlamaIndexAgent framework · MIT · #35 overall
Related jobs
Sources for tool facts
mem0ai/mem0 · run-llama/llama_index · Mem0 documentation. Data fetched 2 Nov 2026. Scores and labels are automated, relative and not a hands-on test.
Get the board every Monday
One email a week: the new top 10, the biggest score changes and every safety label that changed.
Free. Double opt-in: you confirm by email first, unsubscribe in one click.