Install once.
Recall through your agent.

1. Install Lore

Windows

Open the Start menu, type PowerShell, and open it. Run these two lines:

Invoke-WebRequest https://raw.githubusercontent.com/fayzkk889/lore/main/install.ps1 -OutFile install-lore.ps1
.\install-lore.ps1

The installer downloads the matching release and verifies its checksum. Close and reopen PowerShell after installation so it can find lore.

Prefer a direct download? Get the Windows x64 ZIP or Windows ARM64 ZIP, with SHA-256 checksums. Extract it, open PowerShell in the extracted folder, and run .\lore.exe connect. The normal installer also adds Lore to your PATH.

2. Connect your coding agent

In your terminal, run:

lore connect

Lore connects the supported clients it detects. Restart your client afterward. To connect one client explicitly, run only its line below:

lore connect codex
lore connect claude-code
lore connect qwen
lore connect cursor
lore connect claude-desktop

If no client is detected, use one of the explicit client names above. Connecting Codex, Claude Code or Qwen Code installs managed default global guidance so history requests can discover Lore. Your existing guidance is preserved.

OpenCode requires manual MCP configuration. Ollama requires a client that executes tools; running an Ollama model by itself does not connect it to Lore. See the current client configuration documentation.

3. Ask in your chat box

In a fresh connected chat, try:

Read the relevant history and continue this work.

For an explicit request that names the tool:

Get the relevant project context from Lore.
Through Lore, find why we chose this deployment approach.
Remember this decision in Lore: use the existing payment handler.
Show my saved Lore memories for this project.

Lore discovers accessible local Codex, Claude Code and Qwen Code history during recall. Cursor and Claude Desktop use shared saved or imported memory; their own native chat capture is not automatic. A model must call the tool to retrieve context. Ask it to name missing evidence rather than invent details.

Memory lives locally at ~/.lore/archive.json by default. A different computer needs its own setup and an explicit memory transfer; there is no automatic cloud sync.

Browser project memory — developer preview

1. Install Lore, then download the extension preview ZIP and extract it into a folder.
2. Open chrome://extensions or edge://extensions. Enable Developer mode, choose Load unpacked, and select the extracted folder containing manifest.json.
3. In your terminal, run lore browser install once.
4. Open a saved ChatGPT or Claude conversation, click Lore’s toolbar icon, choose the matching project, and explicitly enable capture for that conversation.
5. In a new browser chat, open Lore, choose the project, and click Get project context. Insert into an empty draft or copy it, then review and send yourself.
6. In your coding agent, ask for relevant history through Lore. Choose an existing workspace in the extension for its default scope, or ask for a named project explicitly.

Capture covers opened, rendered messages rather than all account history. Stable-ID edits replace their saved turns. When a layout lacks message IDs, older edited versions can remain; the popup reports this. Browser timestamps are capture times, not original authored dates. Attachments and complete regenerated branches are not captured.

Updating the installed Lore executable at the registered path updates the memory engine used by the extension. Moving the app requires rerunning lore browser install. This unpacked preview requires manual updates for extension code; store-managed updates come after publication.

If something doesn’t work