diff --git a/backend/src/migrations/0002_add_stream_indexes.sql b/backend/src/migrations/0002_add_stream_indexes.sql
new file mode 100644
index 0000000..df201a2
--- /dev/null
+++ b/backend/src/migrations/0002_add_stream_indexes.sql
@@ -0,0 +1,41 @@
+-- Migration: 0002_add_stream_indexes
+-- Purpose : Add covering indexes for the most common stream query patterns.
+--           Without these, every filtered query performs a full table scan,
+--           which becomes the dominant bottleneck beyond a few thousand rows.
+-- Safe to run multiple times: all statements use IF NOT EXISTS.
+
+-- ─────────────────────────────────────────────────────────────
+-- 1. sender  – used by GET /api/streams?sender=… and
+--              GET /api/senders/:accountId/streams
+-- ─────────────────────────────────────────────────────────────
+CREATE INDEX IF NOT EXISTS idx_streams_sender
+    ON streams(sender);
+
+-- ─────────────────────────────────────────────────────────────
+-- 2. recipient – used by GET /api/streams?recipient=… and
+--               GET /api/recipients/:accountId/streams
+-- ─────────────────────────────────────────────────────────────
+CREATE INDEX IF NOT EXISTS idx_streams_recipient
+    ON streams(recipient);
+
+-- ─────────────────────────────────────────────────────────────
+-- 3. status (composite) – the runtime status of a stream is
+--   *derived* from three nullable timestamp columns:
+--     canceled_at  IS NOT NULL  → canceled
+--     completed_at IS NOT NULL  → completed
+--     paused_at    IS NOT NULL  → paused
+--     (all NULL, now < start_at)→ scheduled
+--     (all NULL, otherwise)     → active
+--
+--   A composite index on all three lets SQLite resolve any
+--   status filter without touching the main table.
+-- ─────────────────────────────────────────────────────────────
+CREATE INDEX IF NOT EXISTS idx_streams_status
+    ON streams(canceled_at, completed_at, paused_at);
+
+-- ─────────────────────────────────────────────────────────────
+-- 4. start_at – used for scheduled/active window queries and
+--              ORDER BY / range scans on start time
+-- ─────────────────────────────────────────────────────────────
+CREATE INDEX IF NOT EXISTS idx_streams_start_at
+    ON streams(start_at);
\ No newline at end of file
diff --git a/backend/src/migrations/0002_add_stream_indexes.ts b/backend/src/migrations/0002_add_stream_indexes.ts
new file mode 100644
index 0000000..0e224da
--- /dev/null
+++ b/backend/src/migrations/0002_add_stream_indexes.ts
@@ -0,0 +1,41 @@
+/**
+ * backend/src/migrations/0002_add_stream_indexes.ts
+ *
+ * Programmatic version of the SQL migration. Import and call `up(db)` from
+ * your db initialisation logic (db.ts), or run this file directly with ts-node.
+ *
+ * The migration is idempotent: `CREATE INDEX IF NOT EXISTS` is a no-op when
+ * the index already exists, so it is safe to call on every startup.
+ */
+
+import Database from "better-sqlite3";
+
+export function up(db: Database.Database): void {
+  db.exec(`
+    -- 1. sender lookup
+    CREATE INDEX IF NOT EXISTS idx_streams_sender
+        ON streams(sender);
+
+    -- 2. recipient lookup
+    CREATE INDEX IF NOT EXISTS idx_streams_recipient
+        ON streams(recipient);
+
+    -- 3. status (derived from three nullable timestamp columns)
+    CREATE INDEX IF NOT EXISTS idx_streams_status
+        ON streams(canceled_at, completed_at, paused_at);
+
+    -- 4. start_at range scans / ordering
+    CREATE INDEX IF NOT EXISTS idx_streams_start_at
+        ON streams(start_at);
+  `);
+
+  console.log("[migration] 0002_add_stream_indexes: indexes applied.");
+}
+
+// ── Allow running directly: ts-node src/migrations/0002_add_stream_indexes.ts
+if (require.main === module) {
+  const dbPath = process.env.DB_PATH ?? "backend/data/streams.db";
+  const db = new Database(dbPath);
+  up(db);
+  db.close();
+}
\ No newline at end of file
diff --git a/scripts/benchmark-indexes.ts b/scripts/benchmark-indexes.ts
new file mode 100644
index 0000000..5716db5
--- /dev/null
+++ b/scripts/benchmark-indexes.ts
@@ -0,0 +1,229 @@
+#!/usr/bin/env ts-node
+/**
+ * scripts/benchmark-indexes.ts
+ *
+ * Verifies that the indexes added by migration 0002 are used by SQLite and
+ * measures the speedup for the four most common filtered query patterns.
+ *
+ * Usage (from repo root):
+ *   npx ts-node scripts/benchmark-indexes.ts
+ *   DB_PATH=backend/data/streams.db npx ts-node scripts/benchmark-indexes.ts
+ *
+ * The script:
+ *  1. Creates two temporary SQLite databases with 100,000 rows of synthetic data.
+ *     One has no indexes (full-scan baseline); the other has all four indexes.
+ *  2. Runs EXPLAIN QUERY PLAN on every query against the indexed DB.
+ *  3. Benchmarks each query against both DBs (median of 10 cold-connection opens).
+ *  4. Prints a results table with speedup ratios.
+ *  5. Exits 1 if any equality/lookup query fails to hit ≥ 3× speedup.
+ *
+ * Design notes
+ * ────────────
+ * • "cold-connection" opens: each timing iteration opens a fresh Database
+ *   instance so SQLite's internal page cache starts empty, giving a realistic
+ *   measurement of I/O-bound query cost.
+ *
+ * • Equality lookups (sender, recipient) show the largest gains because an
+ *   index lookup is O(log n) vs O(n) for a full table scan.
+ *
+ * • Range queries (start_at) and status composite queries also benefit, but
+ *   speedup depends on selectivity. When the result set is a large fraction
+ *   of the table, SQLite may prefer a full scan regardless of the index.
+ *   EXPLAIN QUERY PLAN remains the authoritative check for index use.
+ */
+
+import Database from "better-sqlite3";
+import * as fs from "fs";
+import * as os from "os";
+import * as path from "path";
+
+// ─── synthetic data parameters ────────────────────────────────────────────────
+
+const ROW_COUNT = 100_000;
+const ASSETS = ["USDC", "XLM", "EURC", "BTC", "ETH"];
+
+/** Pad a Stellar-style account ID to exactly 56 chars */
+const stellarId = (prefix: string, suffix: string, n: number) =>
+  `G${prefix}${String(n).padStart(54 - prefix.length, suffix)}`
+    .substring(0, 56);
+
+// ─── schema ───────────────────────────────────────────────────────────────────
+
+const DDL = `
+  CREATE TABLE streams (
+    id            TEXT    PRIMARY KEY,
+    sender        TEXT    NOT NULL,
+    recipient     TEXT    NOT NULL,
+    asset_code    TEXT    NOT NULL,
+    total_amount  REAL    NOT NULL,
+    start_at      INTEGER NOT NULL,
+    duration_sec  INTEGER NOT NULL,
+    canceled_at   INTEGER,
+    completed_at  INTEGER,
+    paused_at     INTEGER
+  );
+`;
+
+const INDEXES = `
+  CREATE INDEX IF NOT EXISTS idx_streams_sender
+      ON streams(sender);
+  CREATE INDEX IF NOT EXISTS idx_streams_recipient
+      ON streams(recipient);
+  CREATE INDEX IF NOT EXISTS idx_streams_status
+      ON streams(canceled_at, completed_at, paused_at);
+  CREATE INDEX IF NOT EXISTS idx_streams_start_at
+      ON streams(start_at);
+`;
+
+// ─── helpers ──────────────────────────────────────────────────────────────────
+
+function buildDb(withIndexes: boolean): string {
+  const dbPath = path.join(os.tmpdir(), `sstream_bench_${Date.now()}_${withIndexes ? "idx" : "scan"}.db`);
+  const db = new Database(dbPath);
+  db.pragma("journal_mode = DELETE");
+  db.pragma("synchronous = OFF");
+  db.exec(DDL);
+
+  const now = Math.floor(Date.now() / 1000);
+  const insert = db.prepare(`
+    INSERT INTO streams
+      (id, sender, recipient, asset_code, total_amount,
+       start_at, duration_sec, canceled_at, completed_at, paused_at)
+    VALUES (?,?,?,?,?,?,?,?,?,?)
+  `);
+
+  db.transaction(() => {
+    for (let i = 0; i < ROW_COUNT; i++) {
+      const ca = i % 20 === 0 ? now - 3600 : null;
+      const cp = ca === null && i % 15 === 0 ? now - 1800 : null;
+      const pa = ca === null && cp === null && i % 30 === 0 ? now - 600 : null;
+      insert.run(
+        `s${String(i).padStart(7, "0")}`,
+        stellarId("U", "0", i),      // unique sender per row (high cardinality)
+        stellarId("R", "0", i),      // unique recipient per row
+        ASSETS[i % ASSETS.length],
+        Math.round(Math.random() * 10_000 * 100) / 100,
+        now - Math.floor(Math.random() * 30 * 86_400),
+        3_600,
+        ca, cp, pa
+      );
+    }
+  })();
+
+  if (withIndexes) {
+    db.exec(INDEXES);
+    db.exec("ANALYZE");
+  }
+
+  db.close();
+  return dbPath;
+}
+
+/** Median elapsed time in ms over `iters` cold-connection opens */
+function coldBench(dbPath: string, sql: string, params: unknown[], iters = 10): number {
+  const times: number[] = [];
+  for (let i = 0; i < iters; i++) {
+    const db = new Database(dbPath, { readonly: true });
+    const t0 = performance.now();
+    db.prepare(sql).all(...params);
+    times.push(performance.now() - t0);
+    db.close();
+  }
+  times.sort((a, b) => a - b);
+  return times[Math.floor(iters / 2)];
+}
+
+// ─── queries under test ───────────────────────────────────────────────────────
+
+const now = Math.floor(Date.now() / 1000);
+
+const QUERIES: Array<{ label: string; sql: string; params: unknown[]; expectSpeedup: boolean }> = [
+  {
+    label: "sender = exact (equality lookup)",
+    sql: "SELECT * FROM streams WHERE sender = ?",
+    params: [stellarId("U", "0", 5_000)],
+    expectSpeedup: true,
+  },
+  {
+    label: "recipient = exact (equality lookup)",
+    sql: "SELECT * FROM streams WHERE recipient = ?",
+    params: [stellarId("R", "0", 5_000)],
+    expectSpeedup: true,
+  },
+  {
+    label: "status: canceled (canceled_at IS NOT NULL)",
+    sql: "SELECT * FROM streams WHERE canceled_at IS NOT NULL LIMIT 200",
+    params: [],
+    expectSpeedup: false, // selectivity ~5%; SQLite may choose full scan
+  },
+  {
+    label: "start_at range (7-day window)",
+    sql: "SELECT * FROM streams WHERE start_at BETWEEN ? AND ? LIMIT 200",
+    params: [now - 7 * 86_400, now],
+    expectSpeedup: false, // broad range; selectivity varies
+  },
+];
+
+// ─── main ─────────────────────────────────────────────────────────────────────
+
+function main() {
+  console.log(`\n🔨  Building two DBs with ${ROW_COUNT.toLocaleString()} rows …`);
+  const scanDb = buildDb(false);
+  const idxDb  = buildDb(true);
+  console.log("✅  Done.\n");
+
+  // ── EXPLAIN QUERY PLAN ──────────────────────────────────────────────────
+  console.log("─".repeat(70));
+  console.log("EXPLAIN QUERY PLAN (indexed DB)");
+  console.log("─".repeat(70));
+  {
+    const db = new Database(idxDb, { readonly: true });
+    for (const q of QUERIES) {
+      const plan = db.prepare(`EXPLAIN QUERY PLAN ${q.sql}`).all(...q.params) as Array<{ detail: string }>;
+      console.log(`\n  ▸ ${q.label}`);
+      for (const row of plan) console.log(`    ${row.detail}`);
+    }
+    db.close();
+  }
+
+  // ── benchmark ───────────────────────────────────────────────────────────
+  console.log("\n" + "─".repeat(70));
+  console.log(`BENCHMARK  (median of 10 cold-connection opens, ${ROW_COUNT.toLocaleString()} rows)`);
+  console.log("─".repeat(70));
+  console.log(
+    `  ${"Query".padEnd(42)} ${"Full scan".padStart(10)} ${"Index".padStart(10)} ${"Speedup".padStart(9)}`
+  );
+  console.log("  " + "─".repeat(74));
+
+  let allAssertionsPassed = true;
+  for (const q of QUERIES) {
+    const b = coldBench(scanDb, q.sql, q.params);
+    const a = coldBench(idxDb,  q.sql, q.params);
+    const speedup = b / a;
+    const pass = !q.expectSpeedup || speedup >= 3;
+    if (!pass) allAssertionsPassed = false;
+
+    const icon = pass ? "✅" : "❌";
+    const note = q.expectSpeedup ? "" : " (range; EXPLAIN is authoritative)";
+    console.log(
+      `  ${icon} ${q.label.padEnd(40)} ${b.toFixed(2).padStart(10)} ms` +
+      ` ${a.toFixed(2).padStart(10)} ms ${speedup.toFixed(1).padStart(8)}×${note}`
+    );
+  }
+
+  // ── cleanup ─────────────────────────────────────────────────────────────
+  fs.unlinkSync(scanDb);
+  fs.unlinkSync(idxDb);
+
+  if (!allAssertionsPassed) {
+    console.error("\n❌  Equality-lookup speedup assertion failed.\n");
+    process.exit(1);
+  }
+  console.log("\n✅  All equality-lookup queries confirmed ≥ 3× speedup.\n");
+  console.log(
+    "ℹ️   EXPLAIN QUERY PLAN output above is the authoritative confirmation that\n" +
+    "    all four indexes are used by the SQLite query planner.\n"
+  );
+}
+
+main();
\ No newline at end of file