- Home
- Skills
- Productivity
- LocalLore — Offline Memory for Claude Code
LocalLore — Offline Memory for Claude Code
Search past Claude Code sessions locally with keyword and semantic search, project filters, and cited evidence.
Free
Works with the AI Agents you already use
LocalLore — Offline Memory for Claude Code
Example session with this skill installed
Search my history from last week. Why did we switch from the third-party auth library to a custom implementation in the 'phoenix' project?
- Read your context and instructions
- Compiled the locallore
I have analyzed the session history for the 'phoenix' project from the past 7 days. I found three relevant sessions where the authentication architecture was discussed and modified.
Authentication Migration Audit: Project Phoenix
| Date | Session ID | Action | Reasoning Provided in Logs |
|---|---|---|---|
| 2023-10-24 | cc-8821 | Initial Discussion | Noted high latency (avg 450ms) in the auth-lib-v2 middleware during peak load. |
| 2023-10-25 | cc-8945 | Prototype Custom Auth | Implemented LocalAuthProvider to eliminate external API dependency and reduce cold-start times. |
| 2023-10-26 | cc-9012 | Final Cutover | Removed auth-lib-v2 dependency. Logs indicate a 60% improvement in request throughput in staging. |
Evidence Summary
The primary driver for the switch was performance. You mentioned in session cc-8821 that the third-party library was causing a bottleneck in the request pipeline. The custom implementation was chosen to provide tighter integration with our existing Redis cache and to lower the overhead of the authentication handshake.
Files Involved
src/auth/provider.ts(Created)src/middleware/auth.ts(Modified)package.json(Dependency removed)
Next steps
- Review the performance benchmarks recorded in the
cc-9012logs. - Verify if the
Rediscache configuration used in the custom provider needs further tuning for production.
Connects securely to your tools. The creator never sees your data.
About this skill
LocalLore helps Claude Code recall earlier coding work: why a design changed, where a topic was discussed, what you worked on during a period, and which files were involved.
It indexes existing Claude Code session history and combines keyword and semantic search through a local Docker runtime and MCP server. The remember skill starts with your wording, prefers the current project, applies project/date/file filters, and retrieves more context only when needed. Answers cite useful session timestamps, projects, and files, distinguish evidence from inference, and say when reliable memory is unavailable.
Requirements: Claude Code with plugin support, Docker with Docker Compose v2, and existing Claude Code sessions (normally ~/.claude/projects). This is a Claude Code plugin with supporting runtime files; the SKILL.md alone is not a standalone search service.
Install from the included checkout using ./scripts/install.sh, then load it with claude --plugin-dir . Alternatively, use the upstream Claude Code marketplace:
- `/plugin` marketplace add elarsaks/local-lore
- `/plugin` install locallore@locallore
- `/reload-plugins`
- `/locallore:setup`
Ask /locallore:remember Why did we choose SQLite for this project? or ask naturally to search LocalLore. First setup builds the container and downloads the pinned embedding model; it requires internet access and may take 30 minutes or longer. Search and embedding inference run locally afterward, without sending conversations to a hosted search or embedding service.
Version 0.3.1. Free and MIT licensed. The repository is a finished, pinned snapshot and is not actively maintained. Session history may be incomplete or superseded.
Source: https://github.com/elarsaks/local-lore Documentation: https://elarsaks.github.io/local-lore/
Claude Code session history only; requires a running Docker runtime and LocalLore MCP server. First setup requires internet access and can take 30 minutes or longer. Retrieval depends on available session history, which can be incomplete or superseded. Version 0.3.1 is a finished, pinned snapshot and is not actively maintained.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
1 use
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 1 month ago
- Free with an Agensi account
Requires
Permissions
Setup uses shell scripts, Docker, environment variables and downloads a pinned embedding model. The runtime reads local Claude Code session history, writes a derived index/configuration, and exposes LocalLore MCP on localhost. Conversations are not sent to hosted search or embedding services.