diff --git a/apps/web/app/api/chat/chat.test.ts b/apps/web/app/api/chat/chat.test.ts
index 494483d30ab..48a23e7e030 100644
--- a/apps/web/app/api/chat/chat.test.ts
+++ b/apps/web/app/api/chat/chat.test.ts
@@ -13,7 +13,17 @@ vi.mock("@/lib/active-runs", () => ({
 
 // Mock workspace module
 vi.mock("@/lib/workspace", () => ({
+  ensureManagedWorkspaceRouting: vi.fn(),
+  getActiveWorkspaceName: vi.fn(() => "default"),
+  resolveActiveAgentId: vi.fn(() => "main"),
   resolveAgentWorkspacePrefix: vi.fn(() => null),
+  resolveOpenClawStateDir: vi.fn(() => "/home/testuser/.openclaw-dench"),
+  resolveWorkspaceDirForName: vi.fn((name: string) =>
+    name === "default"
+      ? "/home/testuser/.openclaw-dench/workspace"
+      : `/home/testuser/.openclaw-dench/workspace-${name}`,
+  ),
+  resolveWorkspaceRoot: vi.fn(() => "/home/testuser/.openclaw-dench/workspace"),
 }));
 
 // Mock web-sessions shared module
@@ -42,7 +52,17 @@ describe("Chat API routes", () => {
       getRunningSessionIds: vi.fn(() => []),
     }));
     vi.mock("@/lib/workspace", () => ({
+      ensureManagedWorkspaceRouting: vi.fn(),
+      getActiveWorkspaceName: vi.fn(() => "default"),
+      resolveActiveAgentId: vi.fn(() => "main"),
       resolveAgentWorkspacePrefix: vi.fn(() => null),
+      resolveOpenClawStateDir: vi.fn(() => "/home/testuser/.openclaw-dench"),
+      resolveWorkspaceDirForName: vi.fn((name: string) =>
+        name === "default"
+          ? "/home/testuser/.openclaw-dench/workspace"
+          : `/home/testuser/.openclaw-dench/workspace-${name}`,
+      ),
+      resolveWorkspaceRoot: vi.fn(() => "/home/testuser/.openclaw-dench/workspace"),
     }));
     vi.mock("@/app/api/web-sessions/shared", () => ({
       getSessionMeta: vi.fn(() => undefined),
@@ -115,6 +135,40 @@ describe("Chat API routes", () => {
       expect(startRun).toHaveBeenCalled();
     });
 
+    it("maps partial tool output into AI SDK preliminary output chunks", async () => {
+      const { hasActiveRun, subscribeToRun } = await import("@/lib/active-runs");
+      vi.mocked(hasActiveRun).mockReturnValue(false);
+      vi.mocked(subscribeToRun).mockImplementation(((_sessionId, callback) => {
+        callback({
+          type: "tool-output-partial",
+          toolCallId: "tool-1",
+          output: { text: "partial output" },
+        } as never);
+        callback(null);
+        return () => {};
+      }) as never);
+
+      const { POST } = await import("./route.js");
+      const req = new Request("http://localhost/api/chat", {
+        method: "POST",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify({
+          messages: [
+            { id: "m1", role: "user", parts: [{ type: "text", text: "hello" }] },
+          ],
+          sessionId: "s1",
+        }),
+      });
+      const res = await POST(req);
+      const body = await res.text();
+
+      expect(body).toContain('"type":"tool-output-available"');
+      expect(body).toContain('"toolCallId":"tool-1"');
+      expect(body).toContain('"preliminary":true');
+      expect(body).toContain('"text":"partial output"');
+      expect(body).not.toContain("tool-output-partial");
+    });
+
     it("does not reuse an old run when sessionId is absent", async () => {
       const { startRun, hasActiveRun, subscribeToRun, persistUserMessage } = await import("@/lib/active-runs");
       vi.mocked(hasActiveRun).mockReturnValue(true);
@@ -161,6 +215,48 @@ describe("Chat API routes", () => {
       expect(persistUserMessage).toHaveBeenCalledWith("s1", expect.objectContaining({ id: "m1" }));
     });
 
+    it("repairs managed workspace routing before starting a persisted session run", async () => {
+      const { ensureManagedWorkspaceRouting } = await import("@/lib/workspace");
+      const { getSessionMeta } = await import("@/app/api/web-sessions/shared");
+      const { startRun, hasActiveRun, subscribeToRun } = await import("@/lib/active-runs");
+      vi.mocked(hasActiveRun).mockReturnValue(false);
+      vi.mocked(subscribeToRun).mockReturnValue(() => {});
+      vi.mocked(getSessionMeta).mockReturnValue({
+        id: "s1",
+        title: "Chat",
+        createdAt: 1,
+        updatedAt: 1,
+        messageCount: 1,
+        workspaceName: "default",
+        workspaceRoot: "/home/testuser/.openclaw-dench/workspace",
+        workspaceAgentId: "main",
+        chatAgentId: "chat-slot-main-2",
+      } as never);
+
+      const { POST } = await import("./route.js");
+      const req = new Request("http://localhost/api/chat", {
+        method: "POST",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify({
+          messages: [
+            { id: "m1", role: "user", parts: [{ type: "text", text: "repair routing" }] },
+          ],
+          sessionId: "s1",
+        }),
+      });
+      await POST(req);
+      expect(ensureManagedWorkspaceRouting).toHaveBeenCalledWith(
+        "default",
+        "/home/testuser/.openclaw-dench/workspace",
+        { markDefault: false },
+      );
+      expect(startRun).toHaveBeenCalledWith(
+        expect.objectContaining({
+          overrideAgentId: "chat-slot-main-2",
+        }),
+      );
+    });
+
     it("resolves workspace file paths in message", async () => {
       const { resolveAgentWorkspacePrefix } = await import("@/lib/workspace");
       vi.mocked(resolveAgentWorkspacePrefix).mockReturnValue("workspace");
diff --git a/apps/web/app/api/chat/route.ts b/apps/web/app/api/chat/route.ts
index f138ec0505f..e5443438ce3 100644
--- a/apps/web/app/api/chat/route.ts
+++ b/apps/web/app/api/chat/route.ts
@@ -1,5 +1,13 @@
 import type { UIMessage } from "ai";
-import { resolveAgentWorkspacePrefix } from "@/lib/workspace";
+import {
+	resolveActiveAgentId,
+	resolveAgentWorkspacePrefix,
+	resolveOpenClawStateDir,
+	resolveWorkspaceDirForName,
+	resolveWorkspaceRoot,
+	getActiveWorkspaceName,
+	ensureManagedWorkspaceRouting,
+} from "@/lib/workspace";
 import {
 	startRun,
 	startSubscribeRun,
@@ -14,7 +22,6 @@ import {
 import { trackServer } from "@/lib/telemetry";
 import { existsSync, readFileSync } from "node:fs";
 import { join } from "node:path";
-import { resolveOpenClawStateDir } from "@/lib/workspace";
 import { getSessionMeta } from "@/app/api/web-sessions/shared";
 
 export const runtime = "nodejs";
@@ -40,6 +47,22 @@ function deriveSubagentInfo(sessionKey: string): { parentSessionId: string; task
 	return null;
 }
 
+function normalizeLiveStreamEvent(event: SseEvent): SseEvent {
+	// AI SDK's UI stream schema does not define `tool-output-partial`.
+	// It expects repeated `tool-output-available` chunks with
+	// `preliminary: true` while the tool is still running.
+	if (event.type === "tool-output-partial") {
+		return {
+			type: "tool-output-available",
+			toolCallId: event.toolCallId,
+			output: event.output,
+			preliminary: true,
+		};
+	}
+
+	return event;
+}
+
 export async function POST(req: Request) {
 	const {
 		messages,
@@ -122,7 +145,19 @@ export async function POST(req: Request) {
 		});
 
 		const sessionMeta = getSessionMeta(sessionId);
-		const effectiveAgentId = sessionMeta?.chatAgentId ?? sessionMeta?.workspaceAgentId;
+		const workspaceName =
+			sessionMeta?.workspaceName
+			?? getActiveWorkspaceName()
+			?? "default";
+		const workspaceRoot =
+			sessionMeta?.workspaceRoot
+			?? resolveWorkspaceRoot()
+			?? resolveWorkspaceDirForName(workspaceName);
+		ensureManagedWorkspaceRouting(workspaceName, workspaceRoot, { markDefault: false });
+		const effectiveAgentId =
+			sessionMeta?.chatAgentId
+			?? sessionMeta?.workspaceAgentId
+			?? resolveActiveAgentId();
 
 		try {
 			startRun({
@@ -168,11 +203,8 @@ export async function POST(req: Request) {
 						try { controller.close(); } catch { /* already closed */ }
 						return;
 					}
-					// Skip custom event types not in the AI SDK v6 data stream schema;
-					// they're only consumed by the reconnection parser (processEvent).
-					if (event.type === "tool-output-partial") {return;}
 					try {
-						const json = JSON.stringify(event);
+						const json = JSON.stringify(normalizeLiveStreamEvent(event));
 						controller.enqueue(encoder.encode(`data: ${json}\n\n`));
 					} catch { /* ignore */ }
 				},
diff --git a/apps/web/app/components/chat-panel.stream-parser.test.ts b/apps/web/app/components/chat-panel.stream-parser.test.ts
index 48cb9999aec..a1651cf9554 100644
--- a/apps/web/app/components/chat-panel.stream-parser.test.ts
+++ b/apps/web/app/components/chat-panel.stream-parser.test.ts
@@ -72,6 +72,33 @@ describe("createStreamParser", () => {
     ]);
   });
 
+  it("keeps partial tool output visible without marking the tool complete", () => {
+    const parser = createStreamParser();
+
+    parser.processEvent({
+      type: "tool-input-start",
+      toolCallId: "tool-1",
+      toolName: "readFile",
+    });
+    parser.processEvent({
+      type: "tool-output-partial",
+      toolCallId: "tool-1",
+      output: { text: "first chunk" },
+    });
+
+    expect(parser.getParts()).toEqual([
+      {
+        type: "dynamic-tool",
+        toolCallId: "tool-1",
+        toolName: "readFile",
+        state: "input-available",
+        input: {},
+        output: { text: "first chunk" },
+        preliminary: true,
+      },
+    ]);
+  });
+
   it("closes reasoning state on reasoning-end to prevent stuck streaming badges", () => {
     const parser = createStreamParser();
 
diff --git a/apps/web/app/components/chat-stream-status.test.ts b/apps/web/app/components/chat-stream-status.test.ts
new file mode 100644
index 00000000000..a02efb156ef
--- /dev/null
+++ b/apps/web/app/components/chat-stream-status.test.ts
@@ -0,0 +1,99 @@
+import type { UIMessage } from "ai";
+import { describe, expect, it } from "vitest";
+import {
+	getStreamActivityLabel,
+	hasAssistantText,
+	isStatusReasoningText,
+} from "./chat-stream-status";
+
+function assistantMessage(parts: UIMessage["parts"]): UIMessage {
+	return {
+		id: "assistant-1",
+		role: "assistant",
+		parts,
+	} as UIMessage;
+}
+
+describe("chat stream status helpers", () => {
+	it("detects status reasoning labels that should stay out of the transcript body", () => {
+		expect(isStatusReasoningText("Preparing response...")).toBe(true);
+		expect(
+			isStatusReasoningText(
+				"Optimizing session context...\nRetrying with compacted context...",
+			),
+		).toBe(true);
+		expect(isStatusReasoningText("Planning the requested changes")).toBe(false);
+	});
+
+	it("keeps the stream activity row visible after assistant text has started", () => {
+		const label = getStreamActivityLabel({
+			loadingSession: false,
+			isReconnecting: false,
+			status: "streaming",
+			hasRunningSubagents: false,
+			lastMessage: assistantMessage([
+				{ type: "text", text: "Drafting the final answer now..." },
+			] as UIMessage["parts"]),
+		});
+
+		expect(label).toBe("Still streaming...");
+		expect(
+			hasAssistantText(
+				assistantMessage([
+					{ type: "text", text: "Drafting the final answer now..." },
+				] as UIMessage["parts"]),
+			),
+		).toBe(true);
+	});
+
+	it("prefers gateway status reasoning over the generic streaming label", () => {
+		const label = getStreamActivityLabel({
+			loadingSession: false,
+			isReconnecting: false,
+			status: "streaming",
+			hasRunningSubagents: false,
+			lastMessage: assistantMessage([
+				{
+					type: "reasoning",
+					text: "Optimizing session context...\nRetrying with compacted context...",
+				},
+			] as UIMessage["parts"]),
+		});
+
+		expect(label).toBe("Optimizing session context... Retrying with compacted context...");
+	});
+
+	it("surfaces the active tool name while a tool call is still running", () => {
+		const label = getStreamActivityLabel({
+			loadingSession: false,
+			isReconnecting: false,
+			status: "streaming",
+			hasRunningSubagents: false,
+			lastMessage: assistantMessage([
+				{
+					type: "dynamic-tool",
+					toolName: "read_file",
+					toolCallId: "tool-1",
+					state: "input-available",
+					input: {},
+				},
+			] as UIMessage["parts"]),
+		});
+
+		expect(label).toBe("Running Read File...");
+	});
+
+	it("shows waiting for subagents as the top-priority active status", () => {
+		const label = getStreamActivityLabel({
+			loadingSession: false,
+			isReconnecting: false,
+			status: "streaming",
+			hasRunningSubagents: true,
+			lastMessage: assistantMessage([
+				{ type: "text", text: "Initial draft is ready." },
+			] as UIMessage["parts"]),
+		});
+
+		expect(label).toBe("Waiting for subagents...");
+	});
+});
diff --git a/apps/web/app/components/chat-stream-status.ts b/apps/web/app/components/chat-stream-status.ts
new file mode 100644
index 00000000000..33a92bc408f
--- /dev/null
+++ b/apps/web/app/components/chat-stream-status.ts
@@ -0,0 +1,195 @@
+import type { UIMessage } from "ai";
+
+export const STREAM_STATUS_REASONING_LABELS = [
+	"Preparing response...",
+	"Optimizing session context...",
+	"Waiting for subagent results...",
+	"Waiting for subagents...",
+] as const;
+
+type ChatStatus = "submitted" | "streaming" | "ready" | "error";
+type MessagePart = UIMessage["parts"][number];
+
+function collapseWhitespace(text: string): string {
+	return text.trim().replace(/\s+/g, " ");
+}
+
+function humanizeToolName(toolName: string): string {
+	const normalized = toolName
+		.replace(/^tool-/, "")
+		.replace(/[_-]+/g, " ")
+		.trim();
+
+	if (!normalized) {
+		return "tool";
+	}
+
+	return normalized.replace(/\b\w/g, (char) => char.toUpperCase());
+}
+
+function resolveToolName(part: MessagePart): string | null {
+	if (part.type === "dynamic-tool") {
+		return typeof part.toolName === "string" ? part.toolName : null;
+	}
+
+	if (!part.type.startsWith("tool-")) {
+		return null;
+	}
+
+	const toolPart = part as {
+		type: string;
+		title?: unknown;
+		toolName?: unknown;
+	};
+
+	if (typeof toolPart.title === "string" && toolPart.title.trim()) {
+		return toolPart.title;
+	}
+	if (typeof toolPart.toolName === "string" && toolPart.toolName.trim()) {
+		return toolPart.toolName;
+	}
+
+	return part.type.replace(/^tool-/, "");
+}
+
+function resolveToolState(part: MessagePart): string | null {
+	if (part.type === "dynamic-tool") {
+		return typeof part.state === "string"
+			? part.state
+			: "input-available";
+	}
+
+	if (!part.type.startsWith("tool-")) {
+		return null;
+	}
+
+	const toolPart = part as {
+		state?: unknown;
+		errorText?: unknown;
+		output?: unknown;
+		result?: unknown;
+	};
+
+	if (typeof toolPart.state === "string") {
+		return toolPart.state;
+	}
+	if (typeof toolPart.errorText === "string" && toolPart.errorText.trim()) {
+		return "error";
+	}
+	if ("result" in toolPart || "output" in toolPart) {
+		return "output-available";
+	}
+
+	return "input-available";
+}
+
+export function hasAssistantText(message: UIMessage | null): boolean {
+	return Boolean(
+		message?.role === "assistant" &&
+		message.parts.some(
+			(part) =>
+				part.type === "text" &&
+				typeof (part as { text?: unknown }).text === "string" &&
+				(part as { text: string }).text.length > 0,
+		),
+	);
+}
+
+export function isStatusReasoningText(text: string): boolean {
+	return STREAM_STATUS_REASONING_LABELS.some((label) =>
+		text.startsWith(label),
+	);
+}
+
+function getLatestStatusReasoning(parts: UIMessage["parts"]): string | null {
+	for (let i = parts.length - 1; i >= 0; i--) {
+		const part = parts[i];
+		if (part.type !== "reasoning") {
+			continue;
+		}
+
+		const text =
+			typeof (part as { text?: unknown }).text === "string"
+				? collapseWhitespace((part as { text: string }).text)
+				: "";
+
+		if (text && isStatusReasoningText(text)) {
+			return text;
+		}
+	}
+
+	return null;
+}
+
+function getRunningToolLabel(parts: UIMessage["parts"]): string | null {
+	for (let i = parts.length - 1; i >= 0; i--) {
+		const part = parts[i];
+		const state = resolveToolState(part);
+		if (!state || state === "output-available" || state === "error") {
+			continue;
+		}
+
+		const toolName = resolveToolName(part);
+		if (!toolName) {
+			continue;
+		}
+
+		if (toolName === "sessions_spawn") {
+			return "Starting subagent...";
+		}
+
+		return `Running ${humanizeToolName(toolName)}...`;
+	}
+
+	return null;
+}
+
+export function getStreamActivityLabel({
+	loadingSession,
+	isReconnecting,
+	status,
+	hasRunningSubagents,
+	lastMessage,
+}: {
+	loadingSession: boolean;
+	isReconnecting: boolean;
+	status: ChatStatus;
+	hasRunningSubagents: boolean;
+	lastMessage: UIMessage | null;
+}): string | null {
+	if (loadingSession) {
+		return "Loading session...";
+	}
+
+	if (isReconnecting) {
+		return "Resuming stream...";
+	}
+
+	if (hasRunningSubagents) {
+		return "Waiting for subagents...";
+	}
+
+	if (lastMessage?.role === "assistant") {
+		const statusReasoning = getLatestStatusReasoning(lastMessage.parts);
+		if (statusReasoning) {
+			return statusReasoning;
+		}
+
+		const runningTool = getRunningToolLabel(lastMessage.parts);
+		if (runningTool) {
+			return runningTool;
+		}
+	}
+
+	if (status === "submitted") {
+		return "Thinking...";
+	}
+
+	if (status === "streaming") {
+		return hasAssistantText(lastMessage)
+			? "Still streaming..."
+			: "Streaming...";
+	}
+
+	return null;
+}