What is shadowgraph?
shadowgraph is a MCP server by LiLara-AI that local-first decision memory for AI agents: stores choices, rejected alternatives and reasons, failed attempts, outcomes, provenance and review conditions. It is listed under Knowledge & memory and has 3 GitHub stars.
| Type | Mcp server |
|---|---|
| Author | LiLara-AI |
| Repository | LiLara-AI/shadowgraph |
| Popularity | 3 GitHub stars |
| Category | Knowledge & memory |
| Security | Low risk (96/100) |
| Rank | #2,612 of 4,418 MCP servers; #279 in Knowledge & memory |
| License | MIT |
| Last commit | 2026-10-13 |
| Runs | Locally |
| Language | TypeScript |
| Platforms | macOS, Windows, Linux |
Is shadowgraph safe to install?
shadowgraph scores 96/100 on the MCPMods security check: Low risk. No checks raised a concern. It passed 5 of 6 checks.
MCP server results come from listing signals: who publishes it, where your data goes and what it can do. The code itself is not scanned yet.
| Check | Result |
|---|---|
| Community-built server | Note |
| Runs on your machine only | Pass |
| No credentials mentioned | Pass |
| Read-only by description | Pass |
| No code execution | Pass |
| Source on GitHub | Pass |
How do I install shadowgraph?
- Find the launch command. The README says how the server starts (npx, uvx, docker or a remote URL) and which API keys it needs.
https://github.com/LiLara-AI/shadowgraph
- Add it to Claude Code. Replace the part after -- with the command from the README. Remote servers use --transport http with a URL.
claude mcp add shadowgraph -- <command from README> claude mcp add --transport http shadowgraph <server-url>
- Check it connected. This lists your MCP servers and their status.
claude mcp list
Or ask Claude to set it up
Paste this into Claude Code and it installs shadowgraph for you:
Add the "shadowgraph" MCP server from https://github.com/LiLara-AI/shadowgraph to Claude Code. Read its README for the launch command and any API keys I need, run claude mcp add, then check that it connects.