LOREGet LoreInstall and connect
Install Lore, run lore connect to connect detected clients, and restart your client. The guide also shows how to connect one client explicitly.
Private project memory · Free Windows beta
New chat.
Same project context.
Already have months of coding chats? Lore finds relevant decisions, constraints, and unfinished work in accessible local Codex, Claude Code, and Qwen Code history, then shows your agent the source messages.
Windows launch · Local storage · No Lore account or API key needed for memory.
Read the relevant history and continue this work.
Lore · relevant project history
We were adding webhook verification.
- Decision
- HMAC-SHA256, key alias checkout-v5.
- Constraint
- Keep the existing payment flow.
- Next
- Test invalid signatures and replay rejection.
Memory stays on your computerWorks through your agent’s toolsRelevant excerpts, not whole transcripts
ONE SETUP. A NATURAL REQUEST.
Less reconstructing.
More continuing.
Use history you already have
Lore discovers accessible local Codex, Claude Code, and Qwen Code history during recall. Those existing chats can help on your first request.
Ask for the history
In a connected client, ask it to read relevant history through Lore. It retrieves a compact context pack with source references.
Default global guidance supports the plain continuation request in Codex, Claude Code, and Qwen Code. Other hosts may need you to name Lore explicitly.
BUILT FOR PROJECTS THAT OUTLAST A CHAT
The context you
keep needing.
Your past engineering discussions are useful on day one. Lore retrieves evidence from chats you already have and keeps it scoped to the project you are working on.
See everyday requests →Why this approach?
Recover the architecture choices and tradeoffs you already discussed.
What needs to stay true?
Retrieve project requirements, commands, and explicit saved instructions.
Where did we leave off?
Bring recent requests and unfinished work into a fresh session.
KNOW WHAT WORKS BEFORE YOU INSTALL
Your tools.
A shared local memory.
| Client | Recall shared Lore memory | Automatic native history |
|---|---|---|
| Codex · Claude Code · Qwen Code | Supported through MCP setup | Accessible local transcripts |
| Cursor · Claude Desktop | Supported through MCP setup | Use saved/imported memory |
| OpenCode | Manual MCP configuration | Use saved/imported memory |
| Ollama | Requires a tool-executing host | Depends on that host |
| Browser ChatGPT · Claude | Extension context insertion | Opened conversations · preview |
MCP support depends on the host executing Lore’s tools. Memory does not automatically sync between computers. Browser layout coverage and model behavior can vary.
BROWSER DEVELOPER PREVIEWBring browser discussions
Explore the preview →Bring browser discussions
into your project memory.
The optional Chrome/Edge extension saves visible messages from conversations you enable, and retrieves project context for a new chat. It uses the installed Lore app’s memory engine.
Unpacked preview. Browser-store publishing and authenticated live-site validation are pending. It doesn’t fetch your entire account history.
A FEW STRAIGHT ANSWERS
Before you
get started.
Is Lore free? Is it open source?
Lore is free during its beta. The current implementation is closed source; public GitHub contains downloads and documentation. Earlier MIT releases retain their original licence.
Does it need an API key or send my chats to Lore’s servers?
Ordinary memory operations run locally and need no Lore API key or hosted memory service. Retrieved context is supplied to the AI client you choose, whose data policies still apply.
Will it remember everything perfectly?
No. Lore retrieves relevant historical evidence, with a bounded context budget and source references. Missing, incomplete or conflicting history still needs careful handling by the agent.
What about Windows warnings?
Beta executables and installers are unsigned. SmartScreen, Smart App Control or workplace policy can warn or block them. Keep protections enabled and report the exact error if installation is blocked.
LET YOUR NEXT CHAT START WITH CONTEXT