LWC CLI
The lwc CLI in Rust: persistent agent memory as a source-grounded wiki on SQLite, for Claude Code, Codex, Cursor and more
- 72 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- The lwc agent install command edits agent configuration: MCP, hooks and instructions, so review the hooks on first run
- The store holds plaintext of sources, and team spaces sync memory to your server
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/lwcDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The lwc agent install command edits agent configuration: MCP, hooks and instructions, so review the hooks on first run
- The store holds plaintext of sources, and team spaces sync memory to your server
This is third-party code. Review the repository files before installing.
What it does
LWC is a command-line utility that gives agents persistent memory tied to sources. The agent itself recalls, maintains and evolves knowledge across sessions, while a human picks sources, states goals and reviews answers or exported Markdown. `lwc` is a machine interface for the agent, and it is driven by hand only when developing or debugging the tool itself. Memory has four layers: immutable source snapshots, a wiki with pages, citations, links and provenance, temporal memory of changes, decisions and unresolved work, and a schema and purpose layer with project rules. The canonical store is SQLite, and Markdown, full-text indexes and optional graphs are rebuilt from it. Without retrieving chunks on every query, the agent recalls the maintained wiki, reasons from sources and writes verified improvements back. No embeddings or vector database are needed. The index covers documents, passages and sentences: the agent starts with a small context and expands the exact span when necessary. The `lwc agent install --yes` command finds installed agents and configures their MCP, skill, hooks and instructions through idempotent adapters. Supported are Claude Code, Cursor, Codex, OpenCode, Hermes, Gemini CLI, Antigravity, Kiro, GitHub Copilot in VS Code, Copilot CLI, Copilot for JetBrains and pi. Version 0.19.5 added team spaces on a self-hosted server with access permissions and memory recovery.
Who it is for. People who work with agents on long projects and knowledge bases and want conclusions with citations to accumulate across sessions.
Good fit when
- An agent needs shared project memory with sources that updates across sessions
- A team needs shared memory on its own server with space permissions and recovery
- A document base should become a wiki with links and citations so every claim leads to a source
- Memory of one project must be packed into an archive and moved to another store with a conservative merge
Not a fit when
- You need a personal notebook for people: `lwc` is built to be driven by an agent, with a manual mode only for developing the tool
- You need to query a ready vector database: the utility uses no embeddings
- You want to edit notes by hand in an app with a polished UI
Example request
Save the project rules and sources on database migration in global memory and next time start from this wiki.Limitations
The utility is meant to work through an agent: routine commands are not run by hand and you ask the agent to call the `using-lwc` skill, usually as `$using-lwc`. A team space needs your own server and sign-in configuration for email, Feishu or GitHub. Full snapshots are costly during sync, so propagation speed depends on storage and workload.
A memory archive holds the plaintext of the chosen wiki, so share it only with a trusted recipient and treat received content as untrusted data. The license is Apache 2.0, the last commit is from October 7, 2026 and the version at the time of the card is 0.19.5.
How to disable. Run `npx skills remove` for the skill or delete the `using-lwc` folder, then the `lwc agent` command with the matching options removes agent configs. The package itself is removed with `npm uninstall --global @i-xor/lwc`, and memory data lives in the SQLite store directory and is deleted by hand.
FAQ
How do I install LWC?
Run `npm install --global @i-xor/lwc`. Homebrew, crates.io, checksum-verified GitHub releases and a local Cargo build are also supported.
How do I connect LWC to an agent?
With `lwc agent install --yes`: it finds supported agents and configures MCP, skill, hooks and instructions. Then check `lwc agent status --target all --location global` and restart the agent.
How is LWC different from RAG?
The agent works with a maintained wiki where pages, citations and history are revised as knowledge changes, and writes improvements back. There are no embeddings or vector database, and retrieval is one of the operations.
How do I install only the skill?
With `npx skills add JanYork/llm-wiki-cli --skill using-lwc -g`. The skill is invoked as `$using-lwc`.
Which agents does LWC work with?
With 12 registered targets: Claude Code, Cursor, Codex, OpenCode, Hermes, Gemini CLI, Antigravity, Kiro, GitHub Copilot in VS Code, Copilot CLI, Copilot for JetBrains and pi.
Discussion
No questions or reviews yet.
Related
CLI toolsThe awiki CLI keeps agent memory in a markdown vault: search before the web and save whatever is worth keeping.
Local memory for AI agents in plain Markdown files: it keeps only what you approve and serves it through MCP
The ee CLI gives agents durable local memory: rules, decisions and anti-patterns with scores and provenance
Shared memory for coding agents: messages in SQLite, old context compacted into a summary tree and MCP tools to search it.
Local memory for coding agents: sessions turn into a git-versioned Markdown wiki and come back through an MCP server.
Agent memory in Rust: forgets the unimportant, merges similar items, flags contradictions, runs locally, over HTTP and via MCP.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| LWC (this card) | no | 72 | ||
| LLM Wiki Memory | no | not required | 130 | agent memory built on the LLM Wiki idea, while this one runs four layers on SQLite with citations and team spaces |
| Mem0 | no | 68,887 | stores memory for an application through a service, while this one keeps a wiki with sources locally and connects to agents through MCP |
README in short
The description presents LWC as proactive memory for agents and explains how it differs from queries over document chunks. It then covers team spaces, installation, the companion `using-lwc` skill, the four-layer architecture, portable archives and the document graph.