diff --git a/src/providers/cursor.ts b/src/providers/cursor.ts
index 0394923..8963d21 100644
--- a/src/providers/cursor.ts
+++ b/src/providers/cursor.ts
@@ -480,9 +480,7 @@ function loadComposerInputTokens(db: SqliteDatabase): Map<string, number> {
   try {
     const rows = db.query<{ composer_id: string; used: number | null; ctx: number | null }>(COMPOSER_TOKENS_QUERY)
     for (const r of rows) {
-      // Distinguish null (field absent) from 0 (valid token count).
-      // `used ?? ctx` would fall through on 0, suppressing a real zero.
-      const tokens = r.used !== null ? r.used : (r.ctx ?? 0)
+      const tokens = r.used ?? r.ctx ?? 0
       if (r.composer_id && tokens > 0) map.set(r.composer_id, tokens)
     }
   } catch {
@@ -641,6 +639,9 @@ function parseBubbles(
 
       let inputTokens = row.input_tokens ?? 0
       let outputTokens = row.output_tokens ?? 0
+      // The conversation's tools/bash attach to the single call that carries its
+      // real input (its first turn), so they are counted exactly once.
+      let creditedHere = false
 
       // Current Cursor leaves tokenCount at {0,0}. Use the conversation's real
       // context size (promptTokenBreakdown) for input, credited once per
@@ -656,6 +657,7 @@ function parseBubbles(
             } else {
               inputTokens = real
               creditedComposers.add(conversationId)
+              creditedHere = true
             }
           } else {
             inputTokens = Math.ceil(textLen / CHARS_PER_TOKEN)
@@ -692,7 +694,7 @@ function parseBubbles(
       const languages = extractLanguages(blobToText(row.code_blocks))
       const hasCode = languages.length > 0
 
-      const agentTurn = agentTools.get(conversationId)
+      const agentTurn = creditedHere ? agentTools.get(conversationId) : undefined
       const cursorTools: string[] = [
         ...(hasCode ? ['cursor:edit', ...languages.map(l => `lang:${l}`)] : []),
         ...(agentTurn?.tools ?? []),
diff --git a/tests/providers/cursor-real-tokens.test.ts b/tests/providers/cursor-real-tokens.test.ts
index c5c18cb..f2e9bac 100644
--- a/tests/providers/cursor-real-tokens.test.ts
+++ b/tests/providers/cursor-real-tokens.test.ts
@@ -198,13 +198,15 @@ describe.skipIf(skipReason !== null)('cursor real context tokens (#575)', () =>
     expect(credited).toBeDefined()
   })
 
-  it('attributes tools and bash commands from agentKv to the correct conversation', async () => {
+  it('attributes aggregated agentKv tools once in a multi-bubble conversation', async () => {
     const composerId = 'eeee1111-2222-3333-4444-555566667777'
     const requestId = 'req-001'
     const dbPath = buildDb((db) => {
       insertComposerData(db, { composerId, totalUsedTokens: 10000 })
       insertBubble(db, { composerId, bubbleUuid: 'b1', type: 1, text: 'do stuff', requestId })
       insertBubble(db, { composerId, bubbleUuid: 'b2', type: 2, text: 'doing stuff', model: 'gpt-5' })
+      insertBubble(db, { composerId, bubbleUuid: 'b3', type: 1, text: 'do more stuff', requestId: 'req-002' })
+      insertBubble(db, { composerId, bubbleUuid: 'b4', type: 2, text: 'doing more stuff', model: 'gpt-5' })
       // agentKv with tool calls
       insertAgentKv(db, {
         blobId: 'akv-1', role: 'user',
@@ -228,6 +230,12 @@ describe.skipIf(skipReason !== null)('cursor real context tokens (#575)', () =>
     expect(callWithTools!.tools).toContain('Read')
     expect(callWithTools!.tools).toContain('Shell')
     expect(callWithTools!.bashCommands).toContain('npm test')
+
+    const allTools = calls.flatMap(c => c.tools)
+    const allBashCommands = calls.flatMap(c => c.bashCommands)
+    expect(allTools.filter(t => t === 'Read').length).toBe(1)
+    expect(allTools.filter(t => t === 'Shell').length).toBe(1)
+    expect(allBashCommands.filter(cmd => cmd === 'npm test').length).toBe(1)
   })
 
   it('uses conversation model for pricing when input is on a user bubble', async () => {