diff --git a/src/services/api-handlers.ts b/src/services/api-handlers.ts
index b558cc9c..74a58b42 100644
--- a/src/services/api-handlers.ts
+++ b/src/services/api-handlers.ts
@@ -1521,6 +1521,19 @@ export async function handleRunTagMigrationBatch(
               .filter((t: string) => t)
           : [];
 
+        // A memory that already has tags and a tags vector is fully migrated:
+        // re-embedding it produces the same vector and changes nothing. This
+        // endpoint walks the whole list one window at a time, and the normal
+        // case is a handful of untagged memories among hundreds that are done,
+        // so skipping the finished ones is the difference between a few
+        // seconds and re-vectorizing every memory on every run. Memories with
+        // tags but no tags vector are still processed so the missing vector is
+        // rebuilt.
+        if (currentTags.length > 0 && m.tags_vector != null) {
+          migrationProgress.processed++;
+          continue;
+        }
+
         if (currentTags.length === 0) {
           const prompt = `Generate 2-4 short technical tags for this memory content. Call the save_tags tool with a "tags" array.\n\n${m.content}`;
           const result = await provider.executeToolCall(
diff --git a/tests/tag-migration-skip-tagged.test.ts b/tests/tag-migration-skip-tagged.test.ts
new file mode 100644
index 00000000..5559636f
--- /dev/null
+++ b/tests/tag-migration-skip-tagged.test.ts
@@ -0,0 +1,190 @@
+import { afterEach, describe, expect, it } from "bun:test";
+import { mkdtempSync, rmSync, writeFileSync } from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+
+const tempDirs: string[] = [];
+
+afterEach(() => {
+  while (tempDirs.length > 0) {
+    const dir = tempDirs.pop();
+    if (dir) rmSync(dir, { recursive: true, force: true });
+  }
+});
+
+const apiHandlersUrl = new URL("../src/services/api-handlers.js", import.meta.url).href;
+const embeddingUrl = new URL("../src/services/embedding.js", import.meta.url).href;
+const vectorUtilsUrl = new URL("../src/services/turso/vector-utils.js", import.meta.url).href;
+const connectionManagerUrl = new URL("../src/services/turso/connection-manager.js", import.meta.url)
+  .href;
+const shardManagerUrl = new URL("../src/services/turso/shard-manager.js", import.meta.url).href;
+const vectorSearchUrl = new URL("../src/services/turso/vector-search.js", import.meta.url).href;
+const readyUrl = new URL("../src/services/turso/ready.js", import.meta.url).href;
+const userPromptManagerUrl = new URL(
+  "../src/services/user-prompt/user-prompt-manager.js",
+  import.meta.url
+).href;
+const loggerUrl = new URL("../src/services/logger.js", import.meta.url).href;
+const factoryUrl = new URL("../src/services/ai/ai-provider-factory.js", import.meta.url).href;
+const providerConfigUrl = new URL("../src/services/ai/provider-config.js", import.meta.url).href;
+const configUrl = new URL("../src/config.js", import.meta.url).href;
+
+function runScenario(scriptBody: string) {
+  const dir = mkdtempSync(join(tmpdir(), "opencode-mem-tag-skip-"));
+  tempDirs.push(dir);
+  const scriptPath = join(dir, "scenario.mjs");
+  const script = `
+import { mock } from "bun:test";
+
+const updateVectorCalls = [];
+const updateTagCalls = [];
+const embedCalls = [];
+
+// "done1" is fully migrated (tags + tags vector); "todo1" still needs tagging.
+const memories = [
+  { id: "done1", content: "already done content", tags: "alpha,beta", tags_vector: new Uint8Array([1, 2, 3]) },
+  { id: "todo1", content: "needs tagging", tags: null, tags_vector: null },
+];
+
+mock.module(${JSON.stringify(embeddingUrl)}, () => ({
+  embeddingService: {
+    isWarmedUp: true,
+    warmup: async () => {},
+    embedWithTimeout: async (text) => {
+      embedCalls.push(text);
+      return new Float32Array([1, 2, 3]);
+    },
+  },
+}));
+
+mock.module(${JSON.stringify(vectorUtilsUrl)}, () => ({
+  formatTagsForEmbedding: (tags) => "Topics: " + tags.join(", "),
+}));
+
+mock.module(${JSON.stringify(connectionManagerUrl)}, () => ({
+  tursoConnectionManager: {
+    getConnection: async () => ({
+      all: async () => memories,
+      run: async (sql, params) => {
+        updateTagCalls.push({ sql, params });
+      },
+      get: async () => ({ count: memories.length }),
+    }),
+    closeAll: async () => {},
+  },
+}));
+
+mock.module(${JSON.stringify(shardManagerUrl)}, () => ({
+  tursoShardManager: {
+    async getAllShards() {
+      return [{ id: 1, scope: "project", scopeHash: "h", shardIndex: 0, dbPath: "/tmp/shard.db" }];
+    },
+    async withScopeWriteLock(_scope, _hash, operation) {
+      return operation();
+    },
+    async getWriteShard() {
+      return { id: 1, scope: "project", scopeHash: "h", shardIndex: 0, dbPath: "/tmp/shard.db" };
+    },
+    async incrementVectorCount() {},
+  },
+}));
+
+mock.module(${JSON.stringify(vectorSearchUrl)}, () => ({
+  tursoVectorSearch: {
+    updateVector: async (_db, id) => {
+      updateVectorCalls.push(id);
+    },
+    insertVector: async () => {},
+    listMemories: async () => [],
+  },
+}));
+
+mock.module(${JSON.stringify(readyUrl)}, () => ({
+  ensureTursoReady: async () => {},
+}));
+
+mock.module(${JSON.stringify(userPromptManagerUrl)}, () => ({
+  userPromptManager: {},
+}));
+
+mock.module(${JSON.stringify(loggerUrl)}, () => ({
+  log: () => {},
+}));
+
+mock.module(${JSON.stringify(configUrl)}, () => ({
+  CONFIG: { memoryProvider: "openai-chat" },
+}));
+
+mock.module(${JSON.stringify(providerConfigUrl)}, () => ({
+  buildMemoryProviderConfig: () => ({}),
+}));
+
+mock.module(${JSON.stringify(factoryUrl)}, () => ({
+  AIProviderFactory: {
+    createProvider() {
+      return {
+        async executeToolCall() {
+          return { success: true, data: { tags: ["generated", "tags"] } };
+        },
+      };
+    },
+  },
+}));
+
+const {
+  handleRunTagMigrationBatch,
+  handleGetTagMigrationProgress,
+} = await import(${JSON.stringify(apiHandlersUrl)});
+
+${scriptBody}
+`;
+  writeFileSync(scriptPath, script, "utf-8");
+  const result = Bun.spawnSync({
+    cmd: [process.execPath, scriptPath],
+    stdout: "pipe",
+    stderr: "pipe",
+  });
+  const stdout = Buffer.from(result.stdout).toString("utf8").trim();
+  const stderr = Buffer.from(result.stderr).toString("utf8").trim();
+  const jsonLine = stdout
+    .split("\n")
+    .reverse()
+    .find((line) => line.trim().startsWith("{"));
+
+  return {
+    exitCode: result.exitCode,
+    stdout,
+    stderr,
+    parsed: jsonLine ? JSON.parse(jsonLine) : null,
+  };
+}
+
+describe("tag migration skips already-migrated memories", () => {
+  it("does not re-vectorize a memory that already has tags and a tags vector", () => {
+    const result = runScenario(`
+const batch = await handleRunTagMigrationBatch(5);
+const progress = await handleGetTagMigrationProgress();
+console.log(JSON.stringify({ batch, progress, updateVectorCalls, updateTagCalls, embedCalls }));
+`);
+
+    expect(result.exitCode).toBe(0);
+    expect(result.parsed).not.toBeNull();
+
+    // Both memories are accounted for, but only the untagged one is rewritten.
+    expect(result.parsed.batch.data.processed).toBe(2);
+    expect(result.parsed.batch.data.hasMore).toBe(false);
+    expect(result.parsed.batch.data.errors).toBe(0);
+
+    // Only the untagged memory is re-vectorized; the finished one is skipped.
+    expect(result.parsed.updateVectorCalls).toEqual(["todo1"]);
+
+    // Its content is embedded once (content) plus once (tags) — never the done memory.
+    expect(result.parsed.embedCalls).toContain("needs tagging");
+    expect(result.parsed.embedCalls).not.toContain("already done content");
+    expect(result.parsed.embedCalls.length).toBe(2);
+
+    // Only the untagged memory's tags are written to the database.
+    expect(result.parsed.updateTagCalls.length).toBe(1);
+    expect(result.parsed.updateTagCalls[0].params[1]).toBe("todo1");
+  });
+});