diff --git a/export_jmeter.py b/export_jmeter.py
new file mode 100644
index 0000000..b0a28e8
--- /dev/null
+++ b/export_jmeter.py
@@ -0,0 +1,226 @@
+# SPDX-FileCopyrightText: 2023-present Plane Software, Inc.
+# SPDX-License-Identifier: LicenseRef-Plane-Commercial
+#
+# Licensed under the Plane Commercial License (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+# https://plane.so/legals/eula
+#
+# DO NOT remove or modify this notice.
+# NOTICE: Proprietary and confidential. Unauthorized use or distribution is prohibited.
+
+"""
+Export a JMeter-ready CSV (email, password, ws_slug, project_id, state_id,
+member_id) for seeded users in a workspace.
+
+member_id is the user's own User id — the value the work-item `assignee_ids`
+field expects. The JMeter flows pool these across rows (in a setUp step) and
+randomly pick assignees from the CSV instead of calling GET /members/.
+
+Rows are emitted round-robin by project (one row per project in turn) so any
+prefix of N rows covers N distinct projects. The flows bind one row per VU
+sequentially, so this keeps N VUs spread across N projects and avoids funneling
+concurrent work-item creates onto one project's per-project advisory lock.
+
+Use this after `seed_project_at_scale` has run, when you need a CSV that maps
+each user to a workspace + project + state the JMeter test plan can log in to.
+Reads only the DB — does NOT touch users/projects/issues.
+
+Example:
+  python manage.py export_jmeter_csv \\
+    --workspace_slug seedws1 \\
+    --password <password printed by the seed run> \\
+    --output /code/jmeter-seedws1.csv
+"""
+
+# Python imports
+import csv
+import os
+import random
+from collections import defaultdict
+
+# Django imports
+from django.core.management.base import BaseCommand, CommandError
+
+# Module imports
+from plane.db.models import Project, ProjectMember, State, Workspace
+
+
+class Command(BaseCommand):
+    help = (
+        "Export a JMeter-ready CSV (email,password,ws_slug,project_id,state_id,member_id) "
+        "from existing seeded data. Default: ONE row per user (random project + "
+        "state picked per user) — avoids the cartesian-product blow-up and "
+        "guarantees a JMeter thread can't collide with itself on the same email. "
+        "Pass --rows_per_user 0 to emit the full (user × project) cartesian, or "
+        "any positive N to sample N projects per user."
+    )
+
+    def add_arguments(self, parser):
+        parser.add_argument("--workspace_slug", type=str, required=True)
+        parser.add_argument(
+            "--password",
+            type=str,
+            required=True,
+            help="Plaintext password written to every row (must match what was used at seed time).",
+        )
+        parser.add_argument(
+            "--project_identifier",
+            type=str,
+            default="",
+            help="Restrict to one project. Default: all projects in the workspace.",
+        )
+        parser.add_argument(
+            "--email_filter",
+            type=str,
+            default="@seed.plane.so",
+            help="Substring match on user.email. Default: '@seed.plane.so' (filters to seeded users).",
+        )
+        parser.add_argument(
+            "--state_name",
+            type=str,
+            default="Todo",
+            help="Name of the state to write into state_id. Pass 'random' to pick per row. Default: Todo.",
+        )
+        parser.add_argument(
+            "--output",
+            type=str,
+            default="",
+            help=(
+                "CSV output path. Defaults to ./jmeter-<workspace_slug>.csv "
+                "(or ./jmeter-<ws>-<proj>.csv when --project_identifier is set)."
+            ),
+        )
+        parser.add_argument(
+            "--rows_per_user",
+            type=int,
+            default=1,
+            help=(
+                "Number of rows per user. Default 1 — pick one random project per user, "
+                "guaranteeing each user appears exactly once. Pass 0 to emit the full "
+                "(user × project) cartesian product, or any N > 1 to sample N projects "
+                "per user."
+            ),
+        )
+        parser.add_argument(
+            "--seed",
+            type=int,
+            default=None,
+            help="Seed the RNG so --rows_per_user / random state selection are reproducible.",
+        )
+
+    def handle(self, *args, **options):
+        if options["seed"] is not None:
+            random.seed(options["seed"])
+
+        ws = Workspace.objects.filter(slug=options["workspace_slug"]).first()
+        if not ws:
+            raise CommandError(f"Workspace '{options['workspace_slug']}' not found")
+
+        proj_qs = Project.objects.filter(workspace=ws)
+        if options["project_identifier"]:
+            proj_qs = proj_qs.filter(identifier=options["project_identifier"].strip().upper())
+        projects = list(proj_qs.order_by("identifier"))
+        if not projects:
+            raise CommandError(
+                f"No matching projects in workspace '{ws.slug}'"
+                + (f" with identifier '{options['project_identifier']}'." if options["project_identifier"] else ".")
+            )
+
+        # state lookup: either fixed-by-name or random-per-row
+        state_random = options["state_name"].strip().lower() == "random"
+        state_by_project: dict = {}
+        states_by_project_id: dict = {}
+        for p in projects:
+            sids = list(State.objects.filter(project=p).values_list("id", flat=True))
+            if not sids:
+                raise CommandError(f"Project {p.identifier} has no states; can't pick a state_id.")
+            states_by_project_id[p.id] = sids
+            if not state_random:
+                named = State.objects.filter(project=p, name=options["state_name"]).values_list("id", flat=True).first()
+                state_by_project[p.id] = named or sids[0]
+
+        # Memberships → (email, project_id) pairs.
+        email_filter = options["email_filter"].strip()
+        pm_qs = ProjectMember.objects.filter(
+            workspace=ws,
+            project_id__in=[p.id for p in projects],
+            is_active=True,
+            deleted_at__isnull=True,
+        )
+        if email_filter:
+            pm_qs = pm_qs.filter(member__email__icontains=email_filter)
+        pairs = list(
+            pm_qs.order_by("member__email", "project__identifier").values_list(
+                "member_id", "member__email", "project_id"
+            )
+        )
+        if not pairs:
+            raise CommandError(
+                f"No ProjectMember rows matched (workspace={ws.slug}, email_filter='{email_filter}'). "
+                "Check that the seed has actually run and that --email_filter matches the seeded domain."
+            )
+
+        # Group by user; optionally sample per-user. member_id (the user's own
+        # User id) is constant per email, so track it in a side map.
+        by_user: dict = defaultdict(list)
+        member_id_by_email: dict = {}
+        for member_id, email, pid in pairs:
+            by_user[email].append(pid)
+            member_id_by_email[email] = member_id
+
+        cap = options["rows_per_user"]
+        if cap > 0:
+            for email, pids in by_user.items():
+                if len(pids) > cap:
+                    by_user[email] = random.sample(pids, cap)
+
+        # Default output path
+        out_path = options["output"].strip()
+        if not out_path:
+            if options["project_identifier"]:
+                out_path = os.path.join(
+                    os.getcwd(),
+                    f"jmeter-{ws.slug}-{options['project_identifier'].strip().lower()}.csv",
+                )
+            else:
+                out_path = os.path.join(os.getcwd(), f"jmeter-{ws.slug}.csv")
+
+        # Emit rows round-robin by project so any prefix of N rows covers N DISTINCT
+        # projects. The JMeter flows bind one CSV row per VU sequentially
+        # (shareMode.all), so a grouped-by-user emission funnels the first N VUs'
+        # work-item creates onto a few hot projects -> the per-project
+        # pg_advisory_xact_lock serializes those concurrent creates and they blow
+        # past the 60s client timeout (only write paths fail; reads never lock).
+        # One row per project in turn spreads N VUs across N distinct projects.
+        # (Same ordering as loadtests/scripts/stride_users_csv.py, applied at source.)
+        emails_by_project: dict = defaultdict(list)
+        for email in sorted(by_user.keys()):
+            for pid in by_user[email]:
+                emails_by_project[pid].append(email)
+        project_order = list(emails_by_project.keys())
+        random.shuffle(project_order)  # reproducible when --seed is set
+        pid_queues = [(pid, iter(emails_by_project[pid])) for pid in project_order]
+
+        password = options["password"]
+        n_rows = 0
+        remaining = sum(len(v) for v in emails_by_project.values())
+        with open(out_path, "w", newline="", encoding="utf-8") as f:
+            w = csv.writer(f)
+            w.writerow(["email", "password", "ws_slug", "project_id", "state_id", "member_id"])
+            while remaining:
+                for pid, q in pid_queues:
+                    email = next(q, None)
+                    if email is None:
+                        continue
+                    sid = random.choice(states_by_project_id[pid]) if state_random else state_by_project[pid]
+                    w.writerow([email, password, ws.slug, pid, sid, member_id_by_email[email]])
+                    n_rows += 1
+                    remaining -= 1
+
+        self.stdout.write(
+            self.style.SUCCESS(
+                f"Wrote {n_rows} rows ({len(by_user)} users × {len(projects)} project(s)) "
+                f"to {out_path}"
+            )
+        )