diff --git a/packages/replay/README.md b/packages/replay/README.md
index 6d869cc..71fc2e1 100644
--- a/packages/replay/README.md
+++ b/packages/replay/README.md
@@ -32,6 +32,23 @@ The reducer receives the ordered slice of events delivered so far; the engine
 clamps the result and stays blind to the payload. See
 [docs/adapter-interface.md](./docs/adapter-interface.md) for the full contract.
 
+### Progress mode: a signal over time
+
+`onProgress` turns the clock + reducer into a live "percent complete over elapsed
+time" signal — updates as playback advances, on seek forward, and on seek back:
+
+```js
+const clock = new PlaybackClock(envelope, {}, adapter)
+clock.onProgress(({ position, progress }) => draw(position, progress))
+clock.play()
+```
+
+Fidelity is exact: at playback time `t` the delivered set is precisely the events
+at offset `≤ t`, so the emitted `progress` matches the original run's progress at
+`t`. Updates fire only when the percentage moves (a flag/unflag emits nothing).
+Subscribe before playing to catch every update; call `progress()` for the current
+value at any time.
+
 ## Offsets
 
 Each event fires at its **offset** — its recorded `t` minus the first event's
diff --git a/packages/replay/index.js b/packages/replay/index.js
index e1d19f9..50edf25 100644
--- a/packages/replay/index.js
+++ b/packages/replay/index.js
@@ -87,6 +87,10 @@ export class PlaybackClock {
     this._anchorPosition = 0
     /** @type {Set<(event: Event) => void>} */
     this._handlers = new Set()
+    /** @type {Set<(update: { position: number, progress: number }) => void>} */
+    this._progressHandlers = new Set()
+    /** Last progress value pushed, so unchanged progress (e.g. a flag) is not re-emitted. */
+    this._lastProgress = /** @type {number | null} */ (null)
   }
 
   /** Total playback length in ms (offset of the last event; 0 if empty). */
@@ -135,6 +139,23 @@ export class PlaybackClock {
     return () => this._handlers.delete(handler)
   }
 
+  /**
+   * Progress-mode subscription: receive `{ position, progress }` updates as
+   * playback advances — the "percent complete over elapsed time" signal. Fires
+   * only when the percentage actually changes (so flags/unflags, which don't
+   * advance progress, emit nothing), on play, seek forward, and seek backward.
+   * Requires an adapter with a `progress` reducer; without one it never emits.
+   * Subscribe before playing to catch every update; use {@link progress} for the
+   * current value at any time. Returns an unsubscribe.
+   *
+   * @param {(update: { position: number, progress: number }) => void} handler
+   * @returns {() => void}
+   */
+  onProgress(handler) {
+    this._progressHandlers.add(handler)
+    return () => this._progressHandlers.delete(handler)
+  }
+
   /**
    * Start (or resume) playback from the current position. Any event already due
    * at the current position fires synchronously; the rest are scheduled at their
@@ -203,6 +224,7 @@ export class PlaybackClock {
       this._cursor--
     }
     this._position = t
+    this._emitProgressIfChanged()
   }
 
   /** Deliver an event to all subscribers. */
@@ -210,6 +232,16 @@ export class PlaybackClock {
     for (const handler of this._handlers) handler(record)
   }
 
+  /** Push a progress update to subscribers, but only when the percentage moved. */
+  _emitProgressIfChanged() {
+    if (this._progressHandlers.size === 0) return
+    const progress = this.progress()
+    if (progress === null || progress === this._lastProgress) return
+    this._lastProgress = progress
+    const update = { position: this._position, progress }
+    for (const handler of this._progressHandlers) handler(update)
+  }
+
   /**
    * Deliver anything already due, then arm a timer for the next pending event.
    * Ends playback when the cursor reaches the last event.
diff --git a/packages/replay/test/progress-mode.test.js b/packages/replay/test/progress-mode.test.js
new file mode 100644
index 0000000..ffc135f
--- /dev/null
+++ b/packages/replay/test/progress-mode.test.js
@@ -0,0 +1,184 @@
+// @ts-check
+import { describe, it, expect } from 'vitest'
+import { PlaybackClock } from '@cozy-games/replay'
+import { createMoveLog } from '@cozy-games/move-log'
+
+// A real mnswpr run + its reducer — imported by the TEST via relative paths, so
+// no game dependency enters the replay engine's manifest.
+import { GameSession, MinesweeperRules } from '../../mnswpr/core/index.js'
+import { createProgressReducer } from '../../mnswpr/adapters/replay-progress.js'
+
+/** Deterministic injected scheduler (the injected-clock seam), no vi needed. */
+function fakeScheduler(start = 0) {
+  let now = start
+  let nextId = 1
+  const timers = new Map()
+  return {
+    clock: () => now,
+    setTimeout: (fn, ms) => {
+      const id = nextId++
+      timers.set(id, { at: now + Math.max(0, ms), fn })
+      return id
+    },
+    clearTimeout: (id) => { timers.delete(id) },
+    advance(ms) {
+      const target = now + ms
+      for (;;) {
+        let due = null
+        for (const [id, timer] of timers) {
+          if (timer.at <= target && (due === null || timer.at < due.at)) due = { id, ...timer }
+        }
+        if (!due) break
+        timers.delete(due.id)
+        now = due.at
+        due.fn()
+      }
+      now = target
+    }
+  }
+}
+
+// 3x3, single mine at (0,0). Total safe = 8.
+const layout = () => ({
+  rows: 3,
+  cols: 3,
+  mines: 1,
+  cells: [
+    [{ mine: true, adjacent: 0 }, { mine: false, adjacent: 1 }, { mine: false, adjacent: 0 }],
+    [{ mine: false, adjacent: 1 }, { mine: false, adjacent: 1 }, { mine: false, adjacent: 0 }],
+    [{ mine: false, adjacent: 0 }, { mine: false, adjacent: 0 }, { mine: false, adjacent: 0 }]
+  ],
+  mineLocations: [[0, 0]]
+})
+
+// Drive a real session, recording BOTH the emitted move-events and the ground
+// truth (revealedSafe) after each move — an independent source of truth.
+const board = layout()
+const TOTAL_SAFE = 8
+const session = new GameSession(MinesweeperRules, { state: MinesweeperRules.fromLayout(board), clock: () => nowClock })
+let nowClock = 0
+const emitted = []
+session.onMove(e => emitted.push(e))
+const truthPoints = [] // { offset, revealedSafe } after each move
+
+const script = [
+  { at: 1000, move: { type: 'reveal', r: 0, c: 1 } }, // +1 safe cell
+  { at: 1100, move: { type: 'reveal', r: 1, c: 0 } }, // +1 safe cell
+  { at: 1200, move: { type: 'flag', r: 0, c: 0 } },   // flag the mine — no progress
+  { at: 1300, move: { type: 'reveal', r: 2, c: 2 } }  // floods the rest → all 8
+]
+const baseT = script[0].at
+for (const step of script) {
+  nowClock = step.at
+  session.applyMove(step.move)
+  truthPoints.push({ offset: step.at - baseT, revealedSafe: session.state.revealedSafe })
+}
+
+const records = emitted.map(e => ({ seq: e.seq, t: e.t, event: e }))
+const envelope = createMoveLog(records)
+
+// Ground truth for the mnswpr reducer: revealedSafe / total at a given offset —
+// derived from the session, NOT from the reducer under test.
+function mnswprTruth(offset) {
+  let revealedSafe = 0
+  for (const p of truthPoints) if (p.offset <= offset) revealedSafe = p.revealedSafe
+  return (revealedSafe / TOTAL_SAFE) * 100
+}
+
+// A second, unrelated adapter: percent of events delivered. Ground truth is the
+// count of records at offset <= t.
+const totalEvents = records.length
+const dummyReduce = events => (events.length / totalEvents) * 100
+function dummyTruth(offset) {
+  return (records.filter(r => (r.t - baseT) <= offset).length / totalEvents) * 100
+}
+
+const CASES = [
+  { name: 'mnswpr percent-cleared', adapter: { progress: createProgressReducer(board) }, truth: mnswprTruth },
+  { name: 'dummy percent-of-events', adapter: { progress: dummyReduce }, truth: dummyTruth }
+]
+
+describe.each(CASES)('progress mode — same code path, adapter: $name', ({ adapter, truth }) => {
+  const clamp = t => Math.max(0, Math.min(t, envelope.events[envelope.events.length - 1].t - baseT))
+
+  it('progress() matches the source run at multiple points (via seek)', () => {
+    const clock = new PlaybackClock(envelope, fakeScheduler(), adapter)
+    for (const t of [-10, 0, 50, 100, 150, 200, 250, 300, 400]) {
+      clock.seek(t)
+      expect(clock.progress()).toBeCloseTo(truth(clamp(t)), 5)
+    }
+  })
+
+  it('progress() matches the source run while playing (fake timers)', () => {
+    const s = fakeScheduler()
+    const clock = new PlaybackClock(envelope, s, adapter)
+    const updates = []
+    clock.onProgress(u => updates.push(u))
+    clock.play()
+
+    let last = 0
+    for (const cp of [0, 100, 200, 300]) {
+      s.advance(cp - last)
+      last = cp
+      expect(clock.progress()).toBeCloseTo(truth(cp), 5)
+    }
+    expect(clock.progress()).toBeCloseTo(truth(clock.duration), 5)
+
+    // The pushed signal is non-decreasing during forward play and ends at 100.
+    const vals = updates.map(u => u.progress)
+    expect(vals).toEqual([...vals].sort((a, b) => a - b))
+    expect(vals.at(-1)).toBeCloseTo(truth(clock.duration), 5)
+    // Each emitted update matches ground truth at the position it reports.
+    for (const u of updates) expect(u.progress).toBeCloseTo(truth(u.position), 5)
+  })
+
+  it('seek forward and backward move the progress signal correctly', () => {
+    const clock = new PlaybackClock(envelope, fakeScheduler(), adapter)
+    const vals = []
+    clock.onProgress(u => vals.push(u.progress))
+
+    clock.seek(clock.duration) // forward to the end
+    expect(clock.progress()).toBeCloseTo(truth(clock.duration), 5)
+
+    clock.seek(0) // jump back to the start
+    expect(clock.progress()).toBeCloseTo(truth(0), 5)
+
+    expect(vals[0]).toBeCloseTo(truth(clock.duration), 5) // went up first
+    expect(vals.at(-1)).toBeCloseTo(truth(0), 5)          // then down
+    expect(vals.at(-1)).toBeLessThan(vals[0])
+  })
+})
+
+describe('progress mode — signal behavior', () => {
+  it('does not emit for events that leave progress unchanged (flags)', () => {
+    // mnswpr: the flag at offset 200 must NOT produce a progress update.
+    const s = fakeScheduler()
+    const clock = new PlaybackClock(envelope, s, { progress: createProgressReducer(board) })
+    const updates = []
+    clock.onProgress(u => updates.push(u))
+    clock.play()
+    s.advance(400)
+    // reveals at 0, 100, 300 changed progress; the flag at 200 did not.
+    expect(updates.map(u => u.position)).toEqual([0, 100, 300])
+    expect(updates.map(u => Math.round(u.progress))).toEqual([13, 25, 100])
+  })
+
+  it('emits nothing when no progress adapter is supplied', () => {
+    const clock = new PlaybackClock(envelope, fakeScheduler())
+    const updates = []
+    clock.onProgress(u => updates.push(u))
+    clock.seek(clock.duration)
+    expect(updates).toEqual([])
+    expect(clock.progress()).toBe(null)
+  })
+
+  it('unsubscribe stops progress delivery', () => {
+    const clock = new PlaybackClock(envelope, fakeScheduler(), { progress: dummyReduce })
+    const updates = []
+    const off = clock.onProgress(u => updates.push(u))
+    clock.seek(100)
+    off()
+    clock.seek(300)
+    expect(updates).toHaveLength(1) // only the first jump delivered
+  })
+})