Bring an existing project
Import a ZIP or connect a repository.
Build a map from the full code inventory.
Know what your AI built, reuse what already exists,
and see what still needs a check.
One feature. Two places.
The dashboard and inbox share delivery rules. Follow the connection before changing either screen.
Runtime checks Not run in this demo
Click a feature. Follow its connections.This is an example, not a scan of your project.
Stop rebuilding what you already have. See what to reuse, what a change touches, and what needs a closer look.
Your project files
Illustrative demo · Example result: inbox.tsx → notifications/delivery.ts
Your AI gets a place to start, so it can review existing notification code instead of building a second version.
CoreMitos looks up recorded features and code references. Your connected AI reads the relevant code and decides what to reuse.
New notification received
● Code change receivedIllustrative demo · Runtime check: not run · A review lead, not a confirmed bug.
A change on the dashboard can affect the inbox too. See the related entry point while there is still time to review it.
CoreMitos compares synced changes with known connections and check records. Your AI can inspect the gap and propose a test.
channel = 'dashboard'message = format(event)
send(message)
saveDelivery(event.id)channel = 'inbox'message = format(event)
send(message)
saveDelivery(event.id)Could these parts be shared?
Keep each screen’s behavior in mind.
Illustrative demo · Illustrative pseudocode · Candidates are not proof of equivalent behavior.
Spot possible duplication before another version gets added. Work with your AI to decide which parts should share code.
CoreMitos surfaces supported duplicate candidates and code evidence. Your AI compares their meaning; it does not merge them automatically.
Come back knowing what you asked for, what changed,
and what has actually been checked.
Example · Sep 30, 2026 · UTC
An updated source snapshot includes the dashboard entry point. Receiving code is separate from running it.
Your exact words · 10:24 UTC
“Add notifications to the dashboard.”
AI report · 10:28 UTC
“The inbox is also connected. Its behavior needs a separate check.”
Illustrative code change · 10:29 UTC
inbox → deliveryinbox → delivery
+ dashboard → deliverysrc/notifications/delivery.tsCurrent snapshot · two entry points
Checks to review · 10:31 UTC
“Test notification delivery from both screens.”
A code change is not a passed test.
Current snapshot · Dashboard and inbox reference delivery logic. Both checks are still not run.
Your AI does the coding. CoreMitos keeps the map,
the evidence, and the next question in view.
Illustrative workflow · supported clients and setup required
“Notifications already exist in the inbox. Here are the delivery code and related screens.”
Your AI calls CoreMitos tools to find existing features and related code before making a change.
Updates depend on tool calls and synced code.
Importing files does not mean they are fully classified. A tool connection does not automatically capture every conversation or code change.
Import a ZIP or connect a repository.
Build a map from the full code inventory.
Build a record from your first request.
Add context as your project grows.
No. Connect a supported AI client through MCP so it can read project context and report work to CoreMitos. You keep coding in your AI tool. The available actions depend on the client and your setup.
Yes. Import an existing code snapshot from ZIP or GitHub. CoreMitos inventories the files; your connected AI reviews their roles and classification. Unreviewed, unsupported, and excluded parts remain visible.
No. A feature classification, a suspected issue, an AI report, and a runtime check are different kinds of evidence. A check only confirms the behavior and code version it actually covered.
Upload an updated snapshot or configure supported GitHub synchronization. Changes that have not been uploaded or synced are not yet visible. Changed code can make earlier review evidence out of date.
Website imports are stored in your private project on the server for analysis. Access through MCP is scoped to the authorized project. The public examples on this page do not contain your uploaded code.
This public snapshot is separate from your project and the latest private Kairos review. Original source names and recorded check wording are preserved below.
Make sense of what comes next.