First question
What should it remember?
Memory changes an AI tool from a one-off prompt box into something that may keep context, skills, workspace history, or user-specific recall across sessions.
AI Resources
Agent memory tools for project context
A focused map for memory, recall, skill-memory, and workspace-context tools that can change what an AI system keeps between sessions.
Use it to compare how projects frame agent memory, then open each LifeHubber overview and original source before relying on storage, privacy, setup, or retention details.
Questions to check
Start with the job or constraint that matters now.
These checks frame the source-linked Resources below. They do not rank products or cover every option.
Data path
Check the storage path
The useful differences are in the details: what gets stored, where it runs, how it retrieves context, and what control a user or operator keeps.
Control
Plan the reset button
Before trying a memory layer, check source materials for retention, deletion, access control, hosting, and sensitive-data handling.
Coverage and freshness
Newest LifeHubber addition included here: August 13, 2026
These groups are selective starting points, not a complete directory. The date reflects the newest included Resource’s LifeHubber added date, not a recheck of every linked source. Check the original source for current setup, terms, limits, privacy, access, costs, and behaviour.
Fresh in this topic
Newer Resources already included in this map
1Recently added Resources from the groups below.
Memory and context layers
Projects that help agents remember
10Start here when the important question is what an agent keeps across sessions, how it retrieves context, and how much control remains with the user or operator.
Mem0
mem0ai/mem0
Why it fits this starting point
A library, hosted platform, self-hosted route, and framework integrations expose the choice between implementation control and service convenience.
Acontext
memodb-io/Acontext
Why it fits this starting point
Distilling run learnings into reusable Markdown skill files shows memory becoming portable working instructions rather than retrieved conversation fragments.
TencentDB Agent Memory
TencentCloud/TencentDB-Agent-Memory
Why it fits this starting point
Shared chat memory, reusable skills, document wikis, code graphs, and per-agent access controls show memory managed as team assets rather than personal recall alone.
ReMe
agentscope-ai/ReMe
Why it fits this starting point
Readable Markdown memory with BM25 and wikilinks by default shows a local-first path that does not require an embedding service.
Hindsight
vectorize-io/hindsight
Why it fits this starting point
Learning over time distinguishes adaptation from simply recalling earlier conversations.
CatchMe
HKUDS/CatchMe
Why it fits this starting point
Always-on local activity capture shows a vectorless memory path, while making the local-model versus cloud-provider boundary part of the decision.
cabinet
cabinetai/cabinet
Why it fits this starting point
A local file-based workspace with scheduled jobs and agent memory keeps knowledge, recurring work, and stored context together.
Rowboat
rowboatlabs/rowboat
Why it fits this starting point
An editable Markdown knowledge graph carries work context into email, meetings, browsing, coding, apps, and scheduled agents inside one desktop coworker.
holaOS
holaboss-ai/holaOS
Why it fits this starting point
A shared local-first desktop brings project memory, files, tools, skills, apps, and several coding agents together, while selected models and integrations may still call outside services.
Memvid
memvid/memvid
Why it fits this starting point
A single portable .mv2 file holds documents, embeddings, indexes, metadata, and recovery state when memory needs to move or back up as one unit.
What to explore next
Keep the project usable outside one memory tool
Save the sources, prompts, decisions, checks, and restart notes that matter beyond any chatbot or memory layer.
