diff --git a/src/auto-reply/commands-registry.data.ts b/src/auto-reply/commands-registry.data.ts
index b8a9691a3..d19bd2ede 100644
--- a/src/auto-reply/commands-registry.data.ts
+++ b/src/auto-reply/commands-registry.data.ts
@@ -157,6 +157,13 @@ function buildChatCommands(): ChatCommandDefinition[] {
       description: "Show current status.",
       textAlias: "/status",
     }),
+    defineChatCommand({
+      key: "plan",
+      nativeName: "plan",
+      description: "Interactive project planning (/plan) with saved artifacts.",
+      textAlias: "/plan",
+      acceptsArgs: true,
+    }),
     defineChatCommand({
       key: "plans",
       nativeName: "plans",
diff --git a/src/auto-reply/reply/commands-core.ts b/src/auto-reply/reply/commands-core.ts
index 4f734ab30..93b0212b6 100644
--- a/src/auto-reply/reply/commands-core.ts
+++ b/src/auto-reply/reply/commands-core.ts
@@ -28,6 +28,7 @@ import {
 } from "./commands-session.js";
 import { handlePluginCommand } from "./commands-plugin.js";
 import { handlePlansCommand } from "./commands-plans.js";
+import { handlePlanCommand } from "./commands-plan.js";
 import type {
   CommandHandler,
   CommandHandlerResult,
@@ -44,6 +45,7 @@ const HANDLERS: CommandHandler[] = [
   handleRestartCommand,
   handleTtsCommands,
   handleHelpCommand,
+  handlePlanCommand,
   handlePlansCommand,
   handleCommandsListCommand,
   handleStatusCommand,
diff --git a/src/auto-reply/reply/commands-plan.test.ts b/src/auto-reply/reply/commands-plan.test.ts
new file mode 100644
index 000000000..ed00bb908
--- /dev/null
+++ b/src/auto-reply/reply/commands-plan.test.ts
@@ -0,0 +1,164 @@
+import fs from "node:fs/promises";
+import os from "node:os";
+import path from "node:path";
+
+import { afterAll, beforeAll, describe, expect, it, vi } from "vitest";
+
+import type { ClawdbotConfig } from "../../config/config.js";
+import type { MsgContext } from "../templating.js";
+import { buildCommandContext, handleCommands } from "./commands.js";
+import { parseInlineDirectives } from "./directive-handling.js";
+
+// Mock clack prompts to simulate TUI interaction.
+const clackHoisted = vi.hoisted(() => {
+  let selectCalls = 0;
+  const select = vi.fn(async ({ options, message }: any) => {
+    selectCalls += 1;
+    // First: choose a section (pick first).
+    if (String(message).includes("Choose a section")) {
+      if (selectCalls > 1) return "__review";
+      return options[0].value;
+    }
+    // Second: action selection etc.
+    return options[0].value;
+  });
+  const text = vi.fn(async ({ initialValue }: any) => initialValue ?? "");
+  const confirm = vi.fn(async () => true);
+  const isCancel = vi.fn((v: any) => v === Symbol.for("clack:cancel"));
+  return { select, text, confirm, isCancel };
+});
+
+vi.mock("@clack/prompts", async () => {
+  return {
+    confirm: clackHoisted.confirm,
+    isCancel: clackHoisted.isCancel,
+    select: clackHoisted.select,
+    text: clackHoisted.text,
+  };
+});
+
+const hoisted = vi.hoisted(() => {
+  const runEmbeddedPiAgent = vi.fn(async ({ prompt }: any) => {
+    if (String(prompt).includes("Generate a compact questionnaire")) {
+      return {
+        payloads: [
+          {
+            text: JSON.stringify({
+              goal: "demo",
+              questions: [
+                {
+                  id: "budget",
+                  section: "Constraints",
+                  prompt: "Budget?",
+                  kind: "select",
+                  required: true,
+                  options: ["$", "$$"],
+                },
+                {
+                  id: "deadline",
+                  section: "Timeline",
+                  prompt: "Deadline?",
+                  kind: "text",
+                },
+              ],
+            }),
+          },
+        ],
+      };
+    }
+
+    return {
+      payloads: [{ text: JSON.stringify({ ok: true }) }],
+    };
+  });
+
+  return { runEmbeddedPiAgent };
+});
+
+// llm-task extension dynamically imports embedded runner in src-first/dist-fallback form.
+vi.mock("../../../src/agents/pi-embedded-runner.js", () => ({
+  runEmbeddedPiAgent: hoisted.runEmbeddedPiAgent,
+}));
+vi.mock("../../../agents/pi-embedded-runner.js", () => ({
+  runEmbeddedPiAgent: hoisted.runEmbeddedPiAgent,
+}));
+
+let testWorkspaceDir = os.tmpdir();
+
+beforeAll(async () => {
+  testWorkspaceDir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-plan-"));
+  await fs.writeFile(path.join(testWorkspaceDir, "AGENTS.md"), "# Agents\n", "utf-8");
+});
+
+afterAll(async () => {
+  await fs.rm(testWorkspaceDir, { recursive: true, force: true });
+});
+
+function buildParams(commandBody: string, cfg: ClawdbotConfig, ctxOverrides?: Partial<MsgContext>) {
+  const ctx = {
+    Body: commandBody,
+    CommandBody: commandBody,
+    CommandSource: "cli",
+    CommandAuthorized: true,
+    Provider: "cli",
+    Surface: "cli",
+    ...ctxOverrides,
+  } as MsgContext;
+
+  const command = buildCommandContext({
+    ctx,
+    cfg,
+    isGroup: false,
+    triggerBodyNormalized: commandBody.trim().toLowerCase(),
+    commandAuthorized: true,
+  });
+
+  return {
+    ctx,
+    cfg,
+    command,
+    directives: parseInlineDirectives(commandBody),
+    elevated: { enabled: true, allowed: true, failures: [] },
+    sessionKey: "agent:main:main",
+    workspaceDir: testWorkspaceDir,
+    defaultGroupActivation: () => "mention",
+    resolvedVerboseLevel: "off" as const,
+    resolvedReasoningLevel: "off" as const,
+    resolveDefaultThinkingLevel: async () => undefined,
+    provider: "cli",
+    model: "test-model",
+    contextTokens: 0,
+    isGroup: false,
+  };
+}
+
+describe("/plan TUI", () => {
+  it("creates a plan directory and writes plan.md + answers.json", async () => {
+    // Make TTY true for interactive mode.
+    (process.stdin as any).isTTY = true;
+    (process.stdout as any).isTTY = true;
+
+    const cfg = {
+      commands: { text: true },
+      agents: { defaults: { model: { primary: "openai/mock-1" } } },
+    } as ClawdbotConfig;
+
+    const params = buildParams("/plan plan a trip", cfg);
+    const result = await handleCommands(params);
+
+    expect(result.shouldContinue).toBe(false);
+    expect(result.reply?.text).toContain("Plan saved");
+
+    const plansDir = path.join(testWorkspaceDir, "plans");
+    const entries = await fs.readdir(plansDir, { withFileTypes: true });
+    const dirs = entries.filter((e) => e.isDirectory()).map((e) => e.name);
+    expect(dirs.length).toBeGreaterThan(0);
+
+    const createdDir = path.join(plansDir, dirs[0]);
+    const planMd = await fs.readFile(path.join(createdDir, "plan.md"), "utf-8");
+    const answers = JSON.parse(await fs.readFile(path.join(createdDir, "answers.json"), "utf-8"));
+
+    expect(planMd).toContain("# Plan");
+    expect(Object.keys(answers).length).toBeGreaterThan(0);
+  });
+});
diff --git a/src/auto-reply/reply/commands-plan.ts b/src/auto-reply/reply/commands-plan.ts
new file mode 100644
index 000000000..18853a3ed
--- /dev/null
+++ b/src/auto-reply/reply/commands-plan.ts
@@ -0,0 +1,350 @@
+import fs from "node:fs/promises";
+import path from "node:path";
+
+import { confirm, isCancel, select, text } from "@clack/prompts";
+
+import type { ClawdbotPluginApi } from "../../plugins/types.js";
+import type { CommandHandler } from "./commands-types.js";
+
+// We import the llm-task tool directly (Clawdbot-native) so standalone TUI runs can
+// still use the agent's configured models via the embedded runner.
+import { createLlmTaskTool } from "../../../extensions/llm-task/src/llm-task-tool.js";
+
+type QuestionKind = "text" | "select" | "multiselect" | "confirm";
+
+export type QuestionSpec = {
+  id: string;
+  section: string;
+  prompt: string;
+  kind: QuestionKind;
+  required?: boolean;
+  options?: string[];
+  placeholder?: string;
+};
+
+type QuestionSet = {
+  title?: string;
+  goal: string;
+  questions: QuestionSpec[];
+};
+
+const QUESTIONS_SCHEMA = {
+  type: "object",
+  properties: {
+    title: { type: "string" },
+    goal: { type: "string" },
+    questions: {
+      type: "array",
+      items: {
+        type: "object",
+        properties: {
+          id: { type: "string" },
+          section: { type: "string" },
+          prompt: { type: "string" },
+          kind: { type: "string", enum: ["text", "select", "multiselect", "confirm"] },
+          required: { type: "boolean" },
+          options: { type: "array", items: { type: "string" } },
+          placeholder: { type: "string" },
+        },
+        required: ["id", "section", "prompt", "kind"],
+        additionalProperties: false,
+      },
+    },
+  },
+  required: ["goal", "questions"],
+  additionalProperties: false,
+};
+
+function slugify(input: string): string {
+  return input
+    .toLowerCase()
+    .replace(/[^a-z0-9]+/g, "-")
+    .replace(/^-+|-+$/g, "")
+    .slice(0, 60);
+}
+
+function nowStamp() {
+  const d = new Date();
+  const pad = (n: number) => String(n).padStart(2, "0");
+  return `${d.getFullYear()}-${pad(d.getMonth() + 1)}-${pad(d.getDate())}-${pad(d.getHours())}${pad(
+    d.getMinutes(),
+  )}`;
+}
+
+async function safeReadJson(filePath: string): Promise<any | null> {
+  try {
+    const text = await fs.readFile(filePath, "utf-8");
+    return JSON.parse(text);
+  } catch (err: any) {
+    if (err?.code === "ENOENT") return null;
+    return null;
+  }
+}
+
+async function writeJson(filePath: string, value: any) {
+  await fs.mkdir(path.dirname(filePath), { recursive: true });
+  await fs.writeFile(filePath, JSON.stringify(value, null, 2), "utf-8");
+}
+
+async function writeTextFile(filePath: string, textContent: string) {
+  await fs.mkdir(path.dirname(filePath), { recursive: true });
+  await fs.writeFile(filePath, textContent, "utf-8");
+}
+
+function buildLlmQuestionPrompt(goal: string) {
+  return (
+    `You are helping a user plan a project. Generate a compact questionnaire grouped into sections.\n` +
+    `Return JSON matching the provided schema.\n\n` +
+    `Requirements:\n` +
+    `- Ask only high-signal questions. Prefer multiple choice when the user can pick from common options.\n` +
+    `- Use sections like Goals, Constraints, Inputs, Outputs, Timeline, Risks (as appropriate).\n` +
+    `- Keep it to ~8-15 questions unless the goal clearly needs more.\n` +
+    `- Use stable ids (snake_case).\n\n` +
+    `GOAL:\n${goal}`
+  );
+}
+
+function buildPlanMarkdown(goal: string, answers: Record<string, any>, questions: QuestionSpec[]) {
+  const byId = new Map(questions.map((q) => [q.id, q] as const));
+  const sections = new Map<string, Array<{ q: QuestionSpec; a: any }>>();
+  for (const [id, a] of Object.entries(answers)) {
+    const q = byId.get(id);
+    if (!q) continue;
+    const list = sections.get(q.section) ?? [];
+    list.push({ q, a });
+    sections.set(q.section, list);
+  }
+
+  const lines: string[] = [];
+  lines.push(`# Plan\n\n## Goal\n${goal}\n`);
+  for (const [section, items] of sections) {
+    lines.push(`## ${section}`);
+    for (const { q, a } of items) {
+      const rendered = Array.isArray(a)
+        ? a.join(", ")
+        : typeof a === "boolean"
+          ? a
+            ? "yes"
+            : "no"
+          : String(a);
+      lines.push(`- **${q.prompt}**: ${rendered}`);
+    }
+    lines.push("");
+  }
+
+  return lines.join("\n").trim() + "\n";
+}
+
+function createFakePluginApi(cfg: any, workspaceDir: string): ClawdbotPluginApi {
+  // Minimal stub for llm-task tool; it only relies on `config` and `pluginConfig`.
+  return {
+    id: "local",
+    name: "local",
+    source: "internal",
+    config: cfg,
+    pluginConfig: {},
+    runtime: {} as any,
+    logger: {
+      info() {},
+      warn() {},
+      error() {},
+      debug() {},
+    },
+    registerTool() {},
+    registerHook() {},
+    registerHttpHandler() {},
+    registerChannel() {},
+    registerGatewayMethod() {},
+    registerCli() {},
+    registerService() {},
+    registerProvider() {},
+    registerCommand() {},
+    resolvePath: (input: string) => path.resolve(workspaceDir, input),
+    on() {},
+  } as any;
+}
+
+export const handlePlanCommand: CommandHandler = async (params, allowTextCommands) => {
+  if (!allowTextCommands) return null;
+
+  if (!params.command.commandBodyNormalized.startsWith("/plan")) return null;
+
+  // Only implement interactive plan builder in the local CLI.
+  const isInteractiveCli =
+    params.ctx.CommandSource === "cli" &&
+    Boolean(process.stdin.isTTY) &&
+    Boolean(process.stdout.isTTY);
+  if (!isInteractiveCli) return null;
+
+  const raw =
+    params.ctx.BodyForCommands ??
+    params.ctx.CommandBody ??
+    params.ctx.RawBody ??
+    params.ctx.BodyStripped ??
+    params.ctx.Body ??
+    "/plan";
+
+  const goal = raw.replace(/^\s*\/plan\s*/i, "").trim();
+  if (!goal) {
+    return {
+      shouldContinue: false,
+      reply: { text: "Usage: /plan <goal>" },
+    };
+  }
+
+  const defaultName = `${nowStamp()}-${slugify(goal.split(/\s+/).slice(0, 6).join(" ")) || "plan"}`;
+  const chosenName = await text({
+    message: "Plan name (folder under workspace/plans/)",
+    initialValue: defaultName,
+    validate: (v) => (!v?.trim() ? "Name required" : undefined),
+  });
+  if (isCancel(chosenName)) {
+    return { shouldContinue: false, reply: { text: "Cancelled." } };
+  }
+
+  const planName = slugify(String(chosenName).trim()) || defaultName;
+  const plansDir = path.join(params.workspaceDir, "plans");
+  const planDir = path.join(plansDir, planName);
+  await fs.mkdir(planDir, { recursive: true });
+
+  const metaPath = path.join(planDir, "meta.json");
+  const answersPath = path.join(planDir, "answers.json");
+  const planMdPath = path.join(planDir, "plan.md");
+
+  const existingAnswers = (await safeReadJson(answersPath)) ?? {};
+
+  await writeJson(metaPath, {
+    name: planName,
+    goal,
+    createdAt: new Date().toISOString(),
+    updatedAt: new Date().toISOString(),
+  });
+
+  // Ask the LLM to generate the question set.
+  const api = createFakePluginApi(params.cfg, params.workspaceDir);
+  const tool = createLlmTaskTool(api);
+
+  const llmResult = await tool.execute("llm-task", {
+    prompt: buildLlmQuestionPrompt(goal),
+    input: { goal, existingAnswers },
+    schema: QUESTIONS_SCHEMA,
+  });
+
+  const questionSet = (llmResult as any).details?.json as QuestionSet;
+  const questions = Array.isArray(questionSet?.questions) ? questionSet.questions : [];
+  if (questions.length === 0) {
+    return {
+      shouldContinue: false,
+      reply: { text: "No questions generated (unexpected)." },
+    };
+  }
+
+  // Group by section.
+  const sectionOrder: string[] = [];
+  const bySection = new Map<string, QuestionSpec[]>();
+  for (const q of questions) {
+    const section = q.section?.trim() || "General";
+    if (!bySection.has(section)) sectionOrder.push(section);
+    const list = bySection.get(section) ?? [];
+    list.push(q);
+    bySection.set(section, list);
+  }
+
+  const answers: Record<string, any> = { ...existingAnswers };
+
+  // Iterate sections with a review loop.
+  while (true) {
+    const section = await select({
+      message: "Choose a section",
+      options: [
+        ...sectionOrder.map((s) => ({ label: s, value: s })),
+        { label: "Review + finalize", value: "__review" },
+      ],
+    });
+    if (isCancel(section)) {
+      return { shouldContinue: false, reply: { text: "Cancelled." } };
+    }
+
+    if (section === "__review") break;
+
+    const qs = bySection.get(String(section)) ?? [];
+    for (const q of qs) {
+      const existing = answers[q.id];
+      const required = Boolean(q.required);
+
+      if (q.kind === "confirm") {
+        const res = await confirm({ message: q.prompt, initialValue: Boolean(existing ?? false) });
+        if (isCancel(res)) return { shouldContinue: false, reply: { text: "Cancelled." } };
+        answers[q.id] = Boolean(res);
+      } else if (q.kind === "select") {
+        const opts = (q.options ?? []).map((o) => ({ label: o, value: o }));
+        if (opts.length === 0) {
+          const res = await text({
+            message: q.prompt,
+            initialValue: existing ? String(existing) : "",
+          });
+          if (isCancel(res)) return { shouldContinue: false, reply: { text: "Cancelled." } };
+          const v = String(res).trim();
+          if (required && !v)
+            return { shouldContinue: false, reply: { text: `Missing required answer: ${q.id}` } };
+          answers[q.id] = v;
+        } else {
+          const res = await select({ message: q.prompt, options: opts });
+          if (isCancel(res)) return { shouldContinue: false, reply: { text: "Cancelled." } };
+          answers[q.id] = res;
+        }
+      } else if (q.kind === "multiselect") {
+        // clack multiselect isn't currently imported here to keep deps minimal in this handler.
+        // Fallback to freeform comma-separated input.
+        const res = await text({
+          message: `${q.prompt} (comma-separated)`,
+          initialValue: Array.isArray(existing) ? existing.join(", ") : "",
+          placeholder: q.placeholder,
+        });
+        if (isCancel(res)) return { shouldContinue: false, reply: { text: "Cancelled." } };
+        const arr = String(res)
+          .split(",")
+          .map((s) => s.trim())
+          .filter(Boolean);
+        if (required && arr.length === 0) {
+          return { shouldContinue: false, reply: { text: `Missing required answer: ${q.id}` } };
+        }
+        answers[q.id] = arr;
+      } else {
+        const res = await text({
+          message: q.prompt,
+          initialValue: existing ? String(existing) : "",
+          placeholder: q.placeholder,
+        });
+        if (isCancel(res)) return { shouldContinue: false, reply: { text: "Cancelled." } };
+        const v = String(res).trim();
+        if (required && !v) {
+          return { shouldContinue: false, reply: { text: `Missing required answer: ${q.id}` } };
+        }
+        answers[q.id] = v;
+      }
+
+      await writeJson(answersPath, answers);
+    }
+  }
+
+  const md = buildPlanMarkdown(goal, answers, questions);
+  await writeTextFile(planMdPath, md);
+  await writeJson(metaPath, {
+    name: planName,
+    goal,
+    createdAt: (await safeReadJson(metaPath))?.createdAt ?? new Date().toISOString(),
+    updatedAt: new Date().toISOString(),
+  });
+
+  return {
+    shouldContinue: false,
+    reply: {
+      text:
+        `✅ Plan saved: ${planName}\n` +
+        `- ${path.relative(params.workspaceDir, planMdPath)}\n` +
+        `- ${path.relative(params.workspaceDir, answersPath)}\n\n` +
+        `Use /plans show ${planName} to view it.`,
+    },
+  };
+};