A browser recording is already active. Call browser_record w

Error message

A browser recording is already active. Call browser_record with action="stop" before starting a new recording.

What it means

The recording module tracks a single active RecordingState; startRecording throws if one is already in progress. Recordings are exclusive, so the caller must stop the current one (browser_record action="stop") before starting a new one.

Source

Thrown at packages/core/src/browser/recording/tools.ts:214

// ---------------------------------------------------------------------------
// Lifecycle helpers
// ---------------------------------------------------------------------------

async function startRecording(
  browser: MastraBrowser,
  opts: {
    maxDurationMs?: number;
    everyNthFrame?: number;
    maxWidth?: number;
    maxHeight?: number;
    outputPath?: string;
    threadId?: string;
    outputDir: string;
  },
): Promise<{ id: string; outputPath: string; maxDurationMs: number }> {
  if (active) {
    throw new Error(
      'A browser recording is already active. Call browser_record with action="stop" before starting a new recording.',
    );
  }

  const maxDurationMs = Math.min(
    HARD_CAP_MAX_DURATION_MS,
    Math.max(1_000, opts.maxDurationMs ?? DEFAULT_MAX_DURATION_MS),
  );
  const everyNthFrame = Math.max(1, Math.floor(opts.everyNthFrame ?? DEFAULT_EVERY_NTH_FRAME));
  const maxWidth = Math.max(160, Math.floor(opts.maxWidth ?? DEFAULT_MAX_WIDTH));
  const maxHeight = Math.max(120, Math.floor(opts.maxHeight ?? DEFAULT_MAX_HEIGHT));

  const id = generateRecordingId();
  const outputPath = resolveOutputPath(id, opts.outputDir, opts.outputPath);

  let stream: ScreencastStream;
  try {
    // Bind screencast to the same thread/page the agent is interacting with.

View on GitHub (pinned to 75dd419e61)

Solutions

Call browser_record with action="stop" before starting a new recording
Ensure test cleanup (afterEach) always stops an active recording, even on failure
Avoid running recordings concurrently in the same process/worker
Catch this error on start to detect and recover from a leaked active recording

Example fix

// before
await browser_record({ action: 'start' });
await browser_record({ action: 'start' }); // throws
// after
await browser_record({ action: 'start' });
await browser_record({ action: 'stop' });
await browser_record({ action: 'start' });

Defensive patterns

Strategy: try-catch

Validate before calling

null

Try / catch

try {
  await browser_record({ action: 'start' });
} catch (e) {
  if (e.message.includes('already active')) {
    await browser_record({ action: 'stop' });
    await browser_record({ action: 'start' });
  } else throw e;
}

Prevention

Always pair start with stop, including in test cleanup (afterEach/finally)
Never start a recording while another may be active in the same process
Avoid sharing recording state across concurrent test workers

When it happens

Trigger: Calling browser_record action="start" twice without an intervening "stop"; a previous recording that never stopped (crashed test, forgotten cleanup); parallel test workers sharing the module state.

Common situations: Retry wrappers that re-run a start step after a mid-test failure; concurrent tests in the same process; a stop call that threw earlier leaving state.active set (note: autoStopTimer usually clears it).

Related errors

AI-assisted analysis of mastra-ai/mastra@75dd419e61 (2026-08-30). Data as JSON: /api/errors/bcfc0e2ba085ae11. Report an issue: GitHub.