Try Selvedge in a real project: can you retrieve a decision next session?

[masondelan]

We're looking for a few independent projects to try Selvedge and tell us whether a recorded decision can be found in a later coding session. This invitation is from the Selvedge maintainers and was prepared with AI assistance.

Selvedge is an open-source MCP server and CLI for recording code decisions, their reasons, and rejected approaches. This is a small, voluntary, unpaid pilot in real projects you control. It is not a request for a testimonial.

The workflow is:

1. Install the current release and configure the agent you already use.
2. Record one real decision with its reason—ideally an approach you rejected or rolled back.
3. In a later session, try to retrieve that decision with prior_attempts, blame, or history, or a supported Claude Code session hook.

Connecting the server does not automatically record every decision. A successful retrieval is useful to inspect, but it does not prove that an agent follows the record or that the tool improves productivity. We welcome confusing setup, failed retrievals, and cases where the history was irrelevant or misleading.

To express interest, reply with the agent(s) you already use, a short non-identifying project label, and whether you can try the workflow in a real project. Please also state explicitly whether you agree to Selvedge maintainers storing and reviewing the redacted notes you choose to provide in our internal Notion workspace, using Codex and Grok for analysis. That permission covers those supplied notes only; it does not authorize a public testimonial or additional data collection.

This GitHub discussion and its comments are public. Do not include private repository names, file paths, code, transcripts, credentials, or identifying customer information. Expressing interest does not yet reserve a participant slot; we'll reply with a short checklist after confirming fit. You can stop participating at any time.

Quickstart · Project and source · Feedback guide