Amp by Sourcegraph

Start by asking Amp for a concrete change, not a vague suggestion. From your terminal or inside VS Code, you can point it at a repository and describe the outcome you want—add an endpoint, refactor a module, tighten type checks, or remove dead code. Amp reads the relevant files, reasons through dependencies, and proposes edits you can review and apply. Use it like a teammate: request a plan first, then approve incremental patches, and keep iterating until tests and linters pass.

For day-to-day engineering work, Amp fits best into repeatable workflows. Have it generate a migration, update call sites, and adjust config in one pass, then ask for a focused follow-up: “run through edge cases,” “add unit tests,” or “prepare a rollback note.” When you’re debugging, share an error log and a failing command, then tell Amp to trace the cause and produce a minimal fix. It’s especially useful for multi-file changes where humans lose context—Amp keeps the thread of reasoning and ties edits back to the original goal.

Teams can turn successful runs into reusable patterns. Keep shared threads for common tasks like release prep, dependency upgrades, or security hardening, so others can repeat the same steps with new inputs. Review what’s being used, standardize prompts for recurring work, and audit what changed across the codebase. Because it lives where developers already operate (CLI and editor), adoption is simple: slot Amp into PR preparation, code review support, and “last-mile” cleanup before merge. more