open agent-memory project
MemBukkit
Rent changes from €800 to €950. Ask what is true now—or what was true before.
runnable · separate repository Explore project → Showcases
See how it runs.
Try an open project, inspect a live workspace, or pick a familiar memory case. The tools let you explore directly; the showcases unfold each event, write, and evidence trail as you scroll.
New here? Memseek is the part of an AI application that remembers. Your application appends what happened; Memseek works out what is durably true, keeps every conclusion linked to its evidence, and hands your agent a short briefing instead of a pile of text. Each showcase below opens with the problem it solves.
About the showcase cases. Their people, events, conversations, and outputs are illustrative examples, not live customer records or captured customer activity. The workspace explorer says explicitly when you choose to connect live data.
Projects & tools
Runnable ways to try the memory system or inspect what a workspace contains.
Showcases
Worked memory cases that make each transformation and its evidence visible.
conversation → memory
The agent memory reactor
Watch Alice’s migration rules become facts, project context, and a safer answer.
simulation + runnable example Open case → agents remember events
Generative agents
Follow a party rumor, an overnight reflection, and a correction across three days.
simulation + runnable example Open case → pages → knowledge
An entire brain, declared
Five Markdown pages become a graph, fact index, durable memory, and cited answers.
simulation + runnable example Open case → beliefs challenge themselves
The self-auditing mind
Two reasonable strategies collide, and every belief opens back to its evidence.
simulation + runnable example Open case → clean up memory
dreams × memseek
Merge duplicates, replace stale memory, and review the proposed store as a diff.
design simulation · blueprint Open case → events → current profile
The living profile
Four CRM events become one current profile, with evidence behind every field.
simulation + runnable example Open case → feedback → better playbook
The maintained skill
A bad refund rule produces a cited replacement that only a person can approve.
simulation + runnable example Open case → memory tools for agents
Plug it into an existing agent
Turn a five-item allowlist into safe search, calendar, prompt, and citation tools.
simulation + runnable example Open case →