selvedge
Active·★ 25·MIT·Updated 2026-10-18
Selvedge provides long-term memory for AI-coded codebases, capturing the 'why' behind agent-made changes directly from the agent's context. It acts as a local MCP server, logging structured events to a SQLite database, allowing developers to query the reasoning for code modifications years later.
#AI Agent Memory#Code Provenance#Git Blame for AI#Developer Tooling#LLM Context Management#SQLite#Agent Trace Standard#Code Observability
01
Features
01Live capture of agent reasoning and intent directly from the agent's context.
02Tracks changes at an entity-level (e.g., DB columns, API routes, environment variables) rather than just lines of code.
03Stores all change events and reasoning locally in a durable, queryable SQLite database.
04Provides CLI tools for blaming, diffing, searching, and managing the history of AI-generated code changes.
05Exports and imports change history compatible with the open Agent Trace v0.1.0 standard.
02
Compatibility
Python
Runtime
Verified via docs
SQLite
Database
Verified via docs
Git
Version Control Integration
Verified via docs
Claude Code
AI Agent Runtime
Verified via docs
Cursor
AI Agent Runtime
Verified via docs
Copilot
AI Agent Runtime
Verified via docs
03
Quick start
1
$ pip install selvedge
2
$ cd your-project
3
$ selvedge setup
04
Use cases
↳Understanding the historical 'why' behind schema decisions and migrations in long-term AI-coded projects.
↳Providing an audit trail for AI-generated code changes, essential for compliance (e.g., SOC 2, EU AI Act).
↳Helping solo developers recall the purpose of AI-generated code in their side projects or internal tools.
↳Enforcing checks for prior attempts on entities to prevent agents from re-introducing reverted decisions.
↳Tracking the full scope of a feature or task across multiple entities and commits using changesets.
05
Alternatives
sigmap★ 645
~97% token reduction for AI coding sessions — zero deps, 33 languages, MCP server
keypaste★ 452
A KeePass-compatible password and secrets manager for people who work with AI agents.
gargantua★ 78
Native macOS system cleaner with YAML-driven safety rules, local AI explainability via MLX, and an MCP server for agent-controlled cleanup workflows.
docguard★ 32
The enforcement tool for Canonical-Driven Development (CDD). Audit, generate, and guard your project documentation. Zero dependencies.
world-model-mcp★ 24
A production-ready MCP server that builds a world model for codebases, preventing hallucinations, repeated mistakes, and regressions in Claude Code.
Comments
Log in to leave a comment
No comments yet. Be the first!