Project memory
Your agents remember your decisions and their mistakes.
Verity keeps a knowledge base of your project's decisions and gotchas, in plain Markdown in your repository. Your agent reads it before it starts, and the reviewer reads it before it judges.
Seed
verity initcreates it, andverity memory seedwrites the first notes from your code, README and CLAUDE.md.Inject
When you send a prompt, the notes that match the files and words involved go into the agent's context, within a token budget.
Review
The reviewer walks the graph from the files that changed, so it already knows your decisions when it judges the change.
Reflect
As a task closes, Verity writes a few lasting lessons as new notes. Findings you mark as false positives stop coming back.
The agent drafts, you confirm
When a task is done, at a PR or a “ship it”, the agent drafts one to three cited observations: a decision, a gotcha or a pattern. It shows them to you and saves the version you confirm. Notes you write carry full confidence and are never rewritten for you.
It is just Markdown
Memory is files in your repo, so you read it, edit it and review it in a pull request like any other code. Point Obsidian at the folder if you like, or browse the same graph in the dashboard, with the notes each review cited.
The knowledge base is shared with everyone on the project, so Verity writes derived facts there: paths, decisions, outcomes. The prompts you typed stay out of it.
Knowledge shared across every repository in an organisation is part of Pro and Fleet.
Agents write the code.
Verity keeps it true.
Start on the Free plan. Working with Claude Code today.
npm install -g @codacy/verity-cli && verity initWorks with