ObsolotDeprecation management that ends in migrations, not wiki pages
Most deprecations are announced once and then forgotten. The notice goes into a wiki page, an email thread, or a Slack message, and the migration stalls — because nobody knows where the deprecated thing is still used, who owns removing it, or what the replacement looks like.
Obsolot does one thing well: it takes a deprecation from announcement, through discovery, to migration-complete.
Notices with a lifecycle, not a paragraph
A deprecation notice in Obsolot is a structured record, not prose. It carries a lifecycle state — Draft, Active, Superseded, Completed — markdown documentation, an end-of-life date, and the search patterns that define what the deprecated thing actually looks like in code.
Because the notice owns its search patterns, announcing a deprecation and finding it are the same act. You don't publish a notice and then separately go hunting.
The usage map builds itself
Obsolot runs code-search jobs against your GitHub and GitLab integrations and builds a live map of file-and-line references to every deprecated component. That map updates as your repositories change, so migration progress is measured rather than estimated — per notice, and per repository.
This is the part that usually doesn't happen. A deprecation without a usage map is a request; a deprecation with one is a work list.
A review queue instead of a search problem
An embedded AI agent classifies each discovered reference and proposes concrete fixes. Those proposals land in a migration suggestion review queue where an engineer approves or rejects them.
The point is the inversion: engineers spend their time reviewing proposed changes rather than searching for the things that need changing.
Keeping everyone else in the loop
Notices link out to GitHub Issues and Jira for traceability, outgoing webhooks fire on entity changes so CI/CD and notification pipelines stay in sync, and shields.io-compatible badges let a repository show its deprecation count inline in its README.
See what it costs, or start with one repository.