ML.ai finds every reference before it edits the definition, then lands the rename and each caller together in one diff you review.
Rename getUserById to findUserById everywhere and update its callers.
Searched the project for getUserById4 references in 4 files
Read userService.ts, route.ts and useUser.ts
One definition and three callers.
I will rename them together so nothing is left on the old name.
Saved 4 files
Describe a task or ask a question
and userService.test.ts, +1 −1
for refactorsDevelopers changing shared code that many other files depend on.
A function, type or constant used in more places than anyone remembers.
The same idea written three different ways across the codebase, made consistent in one pass.
The definition and every caller in one diff, so nothing is left half-migrated.
Changing the definition is easy. These four things decide whether the rest of the code still works.
export function getUserById(id)await getUserById(params.id)useSWR(id, getUserById)getUserById("u1")One name, four places it is used
A rename that misses one call site compiles in your head and fails in production. ML.ai searches for every reference before it edits the definition.
Renamed by hand
userService.tsfindUserById
route.tsgetUserById
useUser.tsgetUserById
One diff
userService.tsfindUserById
route.tsfindUserById
useUser.tsfindUserById
Half-migrated versus one diff
The definition and every caller land together, so the codebase is never left between the old name and the new one.
Searched the project for getUserById4 references
userService.tsdefinition
route.tscaller
useUser.tscaller
userService.test.tstest
References, not just a text match
It follows how the function or type is used across the project instead of replacing every matching string, so unrelated names are left alone.
“Renamed the function”
A description versus the real diff
Every edit is a real diff against your files, so review means reading code. A large refactor is still large, but it is all in one place.
You rename a function, then grep for callers by hand, and find out weeks later that a call site you missed has been broken the whole time.
Terminal
$grep -rn getUserById src/
$rename it in userService.ts
$fix callers, one file at a time
$npm test → ReferenceErrorMissed one
A caller you missed shows up after the merge
ML.ai searches the project for every reference before editing the definition, so the rename and every caller it found land together in one diff.
Rename getUserById to findUserById everywhere
Found 4 references in 4 files
general
general · Search · Read · Write diff
Traces references across the project before touching the definition, and lands the rename or restructure and every call site together in one diff.
architect
architect · Read · Plan · Write
Plans the approach first when a refactor is big enough that the order of work is not obvious, before general carries it out.
A refactor is only safe if every caller was found. ML.ai follows references through the project, edits the definition and the callers together, and holds it all behind a diff.
Explorer
general, the default agent, takes the whole task
Rename getUserById to findUserById everywhere and update its callers.
Explorer
Rename getUserById to findUserById everywhere and update its callers.
Searched for getUserById4 references
Read userService.ts, route.ts and useUser.ts
Planned one rename across 4 files
Read the matches in context
Describe a task or ask a question
Explorer
One diff across four files
Describe a task or ask a question
Explorer
One diff across four files
Allowed · the definition and 3 callers saved in one diff
Describe a task or ask a question
Explorer
general, the default agent, takes the whole task
Rename getUserById to findUserById everywhere and update its callers.
Where refactors usually break, and how ML.ai closes each gap.
getUserById usagesWith ML.ai
userService.tsfindUserById
route.tsfindUserById
useUser.tsfindUserById
A function used in a hook, a route and a test, where only two of the three were on your mind when you renamed it.
How handles it
ML.ai searches for every reference before it edits, so the call site you forgot is in the diff.
Old name still in the codeWith ML.ai
userService.tsrenamed
route.tsrenamed
userService.test.tsrenamed
The definition changes and two callers still use the old name, so the build breaks on a branch that looked finished.
How handles it
The definition and every caller are written together as one diff, so the old name is gone everywhere at once.
One diff, every fileWith ML.ai
userService.ts+1 −1
route.ts+1 −1
useUser.ts+1 −1
A wide change reviewed one file at a time, where each file looks fine and the whole is never read.
How handles it
The whole refactor is one diff in front of you, held behind Allow, Always or Deny before anything saves.
The real edges of what ML.ai can do here, stated plainly instead of left for you to discover.
A call built from a string at runtime may not show up the way a static reference does. Worth checking yourself.
Wants to change 4 files
A refactor across many files is still a large diff to review. The Allow, Always or Deny prompt is the same size as for any change.
For a restructure where the approach itself is not clear, plan it with architect, then hand the edit to general.
Apply edits to 3 files?
One diff across the schema, the API route and the task card.
Describe a feature in a sentence and ML.ai writes it across every file it touches, held behind one real diff before anything saves.
Plan · move off the .ini loader
Nothing edited until you hand it to general
explore maps what the old code does today, architect plans the path, and general carries it out as small reviewable diffs.
It searches and reads the project for references before editing, which catches static usages reliably. A call built dynamically from a string at runtime is the case worth double-checking yourself.
The diff shows exactly what changed, so any behavior drift is visible before you approve. A large refactor is still worth reading closely, the same way you would review a teammate’s.
It lands as one coherent diff so the codebase is never left half-migrated, but you can ask for it in stages if you would rather review incrementally.
Yes. Describe the target shape, such as extracting a hook or splitting a file along given lines, and it edits the definition and the callers together.
When the refactor is big enough that the order of work matters, or the target shape is not obvious yet. architect returns the key files and an ordered plan before general edits anything.
today, or talk to us about what is nextInstall the editor agent on your own machine, or book a call to talk through your team’s workloads
