selvedge

Based on the available evidence, selvedge looks like a clearly positioned but very early open-source developer tool. It is not a general vector database or a generic chat-memory add-on; a more accurate analogy is “git blame for why” for AI-made code changes. The current evidence is mostly the project’s own GitHub description, which is enough to infer intent and basic architecture, but not enough to show broad adoption or proven reliability in larger production workflows.

Its practical role is to capture an AI agent’s reasoning, context, and justification while code changes are being made, then store that memory locally in SQLite for later retrieval. For developers repeatedly using coding agents on the same repository, the value is less about generating code directly and more about preserving traceability: why a file was changed, what assumptions were in play, and what context should carry into future edits. So it is better understood as a repository memory/audit layer, not as an IDE, model gateway, or autonomous coding agent itself.

On cost and adoption friction, the evidence only supports that it is open source and mentions local SQLite storage.