Walkthroughs
A walkthrough is an ordered, multi-step explanation of code, pinned across files. Each step is a comment anchored to a specific line; together they tell a story that survives file rewrites.
Walkthroughs are how agents explain code — not just edit it.
Why this is its own primitive
Reading code you didn't write has gotten harder, not easier, as agents do more of the writing. Volume is up. The "author" isn't a teammate you can ping. Patterns shift between runs.
Grove treats understanding code as a first-class operation. Ask an agent to take you through a piece of code, and the result lives in Grove as a navigable artifact — not a wall of chat text — and survives the next refactor, because each step anchors to content rather than line numbers.
Authoring
Agents create walkthroughs through Grove's MCP tools — one call, one walkthrough, as many steps as the story needs. Just ask in chat:
"Walk me through how comment anchoring works. Start at the anchor logic, then show me how the resolver uses it."
Reading
In the Comments panel, a walkthrough renders as a single card with a step counter. Click a step and Grove opens the file, jumps to the anchored line, and highlights it. Each step supports replies and resolution, just like any comment.
Limits
- Agents can author walkthroughs but not retract them yet. You can delete individual step comments to prune.
- Steps anchor to content like comments do, so a heavily rewritten file can leave a step flagged stale — the step still opens the file, but you may need to find the moved line by eye.