Skip to content
Skip to content
Brain Scanner

For developers working with AI coding agents

Stop re-explaining what went wrong to your coding agent.

Brain Scanner keeps the decision trail and change history so you can review and pivot in one place.

Connect Claude Code or Codex through MCP. Review your agent's recorded changes, decisions, checks, blockers, and any submitted file diffs.

Create a free account See a session recovery
Brain Scanner · free

One failed session, recovered

Pivot from the record, not from memory.

The agent's last attempt broke sign-out. Instead of typing the recap again, review what it changed, note your decision, and queue the next step.

BeforeThe recap you keep retyping
  1. You

    No. Last time we moved token refresh into middleware, sign-out broke.

  2. Agent

    Understood. Which files changed in that attempt?

  3. You

    The same ones. And keep the cookie path on /app. We decided that two sessions ago.

  4. Agent

    Got it. Could you paste the failing test output?

Illustrative chat. The context lives in scrollback, so every pivot starts with a recap.

Session reviewExample · runs in your browser · nothing is sent

After · review and pivot in one place

Move token refresh into middleware

Ended with 2 failing checks

Agent summary Moved refresh into request middleware. Sign-out and session-expiry checks now fail.

What broke agent-reported checks

  • logout clears the session cookie
  • expired session redirects to sign-in
  • token refresh extends an active session

What changed 3 files · diffs as submitted

src/routes/logout.ts+2 −3
@@ -8,5 +8,4 @@ export async function logout(req, res) {
-  const session = await sessions.find(req.cookies.sid)
-  if (session) await sessions.revoke(session.id)
-  res.clearCookie('sid', { path: '/app' })
+  // refresh middleware now owns session state
+  return res.redirect('/')
src/auth/middleware.ts+5 −0
@@ -12,0 +12,5 @@
+export async function refreshSession(req, res, next) {
+  const session = await sessions.find(req.cookies.sid)
+  if (session?.expiresSoon) await sessions.extend(session.id)
+  next()
+}
src/auth/session.ts+1 −1
@@ -4,1 +4,1 @@
-const COOKIE_PATH = '/app'
+export const COOKIE_PATH = '/app'

Decisions on record earlier sessions

  • Keep token refresh server-side.
  • The session cookie path stays /app.

Recommended next step from the agent

Restore session revocation in logout.ts, clear the cookie on path /app, then re-run the two failing checks.

Live today: session reviews show agent-submitted summaries, decisions, checks, blockers, and file diffs when provided. Queued tasks can be edited or cancelled from the dashboard. The one-click accept above is an example of that flow.

Inside the workspace

The views behind that recovery, with one example project. Every one is interactive.

    PROJECT⌂   Main Project

    ▷   Example workspace

    PROJECT SNAPSHOT
    Project nodes
    128
    Failed nodes
    0
    Changed or required
    8
    Illustrative values
    Projects › Example workspace ☆INTERACTIVE EXAMPLE

    Explore project structure, recorded context, and queued work.
    Connect Brain Scanner to your coding agent.

    Product interface · illustrative data01 / 08 — Project Graph
    01 / Project Graph

    See how your project
    fits together.

    Your agent maps files, symbols, and subsystems into one connected graph. Follow a relationship to see what sits behind it.

    Select or drag a node, show its relationships, or open its scope.
    02 / Overview

    Know how far along
    the work really is.

    Implementation estimates sit beside their evidence range and confidence, so you can see where work remains.

    Select a capability to inspect its estimate.
    03 / Change Review

    See exactly
    what changed.

    Recorded updates keep their summaries and before-and-after details. Earlier entries stay in history, and published graphs can be compared separately.

    Choose a graph version or open an updated item. This preview contains illustrative data only.
    04 / Context

    Pick up where the
    last session stopped.

    Each session keeps its goal, work, and outcome connected, so the next one starts with history instead of a blank prompt.

    Select a context node. Explore Sessions and Timeline.
    05 / Files

    Every artifact
    has a place.

    Browse mapped folders and files in a familiar explorer, with the path, file type, and recorded details in view.

    Open a folder, follow the breadcrumbs, or select a file to inspect it.
    06 / Reports

    Turn a report
    into the next fix.

    Agents publish reports and link open bugs. Add a bug to the queue, then follow the fix back to its recorded evidence.

    Read an example report and queue its bug. This preview sends nothing to an agent.
    07 / Effort chart

    See where
    effort is visible.

    Resource categories come with their measurement source, so recorded values stay apart from estimates.

    Select a category to inspect its source.
    08 / Queues

    Give the next step
    somewhere to go.

    Scoped tasks wait by type and source until you hand them to your agent. Finished work moves to history.

    Add and remove an example task. Nothing is sent to an agent.
    From your agent to the bigger picture

    Keep working where you are.
    Bring the context with you.

    01Connect your coding agent.Authorize the hosted connector in your agent’s MCP settings.

    02Map a project you can access.Your agent submits structured project intelligence within its permissions.

    03Review. Decide. Queue.Read what the last session changed and why, note your decision, and queue the next step.

    Connect your agent and map a project ↗

    Stop re-explaining.
    Review, decide, and move on.

    Create a free account