LiLara-AI/shadowgraph

Overview

Local-first decision memory for AI agents. It records what was chosen, what was rejected, why, and when to reconsider. Provides MCP, CLI, and HTTP interfaces with zero runtime dependencies.

Best for

Best for
Developers building self-reflective AI agents that need lightweight decision logs

Use cases

• Track agent decision history with reasoning
• Revisit decisions when reconsideration criteria change
• Audit past choices and rejections via MCP or CLI

Notes

Local-first decision memory for AI agents. It records what was chosen, what was rejected, why, and when to reconsider. Provides MCP, CLI, and HTTP interfaces with zero runtime dependencies.

1 stars on GitHub. Last updated 2026-09-04. Licensed MIT.

Use cases

• Track agent decision history with reasoning
• Revisit decisions when reconsideration criteria change
• Audit past choices and rejections via MCP or CLI

Pros

• Local-first design keeps data on device
• Supports MCP, CLI, and HTTP for flexible integration
• No runtime dependencies simplify deployment

Cons

• Single star on GitHub suggests minimal real-world validation
• No active maintenance or version history provided
• Decision memory requires manual integration to be useful

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.