diff --git a/refact-agent/gui/src/components/Chat/Chat.tsx b/refact-agent/gui/src/components/Chat/Chat.tsx
index 6941246b1..5b181e2bf 100644
--- a/refact-agent/gui/src/components/Chat/Chat.tsx
+++ b/refact-agent/gui/src/components/Chat/Chat.tsx
@@ -10,6 +10,7 @@ import {
   useGetCapsQuery,
   useCapsForToolUse,
   useAgentUsage,
+  useLastSentCompressionStop,
 } from "../../hooks";
 import { type Config } from "../../features/Config/configSlice";
 import {
@@ -53,7 +54,7 @@ export const Chat: React.FC<ChatProps> = ({
   const { submit, abort, retryFromIndex } = useSendChatRequest();
 
   const chatToolUse = useAppSelector(getSelectedToolUse);
-
+  const compressionStop = useLastSentCompressionStop();
   const threadNewChatSuggested = useAppSelector(selectThreadNewChatSuggested);
   const messages = useAppSelector(selectMessages);
   const capsForToolUse = useCapsForToolUse();
@@ -105,8 +106,9 @@ export const Chat: React.FC<ChatProps> = ({
         <AgentUsage />
         <SuggestNewChat
           shouldBeVisible={
-            threadNewChatSuggested.wasSuggested &&
-            !threadNewChatSuggested.wasRejectedByUser
+            compressionStop.stopped ||
+            (threadNewChatSuggested.wasSuggested &&
+              !threadNewChatSuggested.wasRejectedByUser)
           }
         />
         {!isStreaming && preventSend && unCalledTools && (
diff --git a/refact-agent/gui/src/components/ChatForm/ChatForm.tsx b/refact-agent/gui/src/components/ChatForm/ChatForm.tsx
index e30d60f9c..eb689f8c6 100644
--- a/refact-agent/gui/src/components/ChatForm/ChatForm.tsx
+++ b/refact-agent/gui/src/components/ChatForm/ChatForm.tsx
@@ -19,7 +19,7 @@ import {
   useSendChatRequest,
   useCompressChat,
   useAutoFocusOnce,
-  useTotalTokenUsage,
+  // useTotalTokenUsage,
 } from "../../hooks";
 import { ErrorCallout, Callout } from "../Callout";
 import { ComboBox } from "../ComboBox";
@@ -35,7 +35,7 @@ import { useInputValue } from "./useInputValue";
 import {
   clearInformation,
   getInformationMessage,
-  setInformation,
+  // setInformation,
 } from "../../features/Errors/informationSlice";
 import { InformationCallout } from "../Callout/Callout";
 import { ToolConfirmation } from "./ToolConfirmation";
@@ -98,18 +98,6 @@ export const ChatForm: React.FC<ChatFormProps> = ({
   const { compressChat, compressChatRequest } = useCompressChat();
   const autoFocus = useAutoFocusOnce();
 
-  const { limitReached, tokens, limit } = useTotalTokenUsage();
-
-  useEffect(() => {
-    if (limitReached) {
-      dispatch(
-        setInformation(
-          `Token Limit reached, ${tokens} out of ${limit} used. To continue click the compress button or start a new chat.`,
-        ),
-      );
-    }
-  }, [tokens, limit, limitReached, dispatch]);
-
   const shouldAgentCapabilitiesBeShown = useMemo(() => {
     return threadToolUse === "agent";
   }, [threadToolUse]);
@@ -131,7 +119,6 @@ export const ChatForm: React.FC<ChatFormProps> = ({
   const disableSend = useMemo(() => {
     // TODO: if interrupting chat some errors can occur
     if (allDisabled) return true;
-    if (limitReached) return true;
     // if (
     //   currentThreadMaximumContextTokens &&
     //   currentThreadUsage?.prompt_tokens &&
@@ -143,7 +130,6 @@ export const ChatForm: React.FC<ChatFormProps> = ({
     return isWaiting || isStreaming || !isOnline || preventSend;
   }, [
     allDisabled,
-    limitReached,
     messages.length,
     isWaiting,
     isStreaming,
@@ -321,11 +307,7 @@ export const ChatForm: React.FC<ChatFormProps> = ({
 
   if (information) {
     return (
-      <InformationCallout
-        mt="2"
-        onClick={onClearInformation}
-        timeout={limitReached ? null : 2000}
-      >
+      <InformationCallout mt="2" onClick={onClearInformation} timeout={2000}>
         {information}
       </InformationCallout>
     );
@@ -385,7 +367,6 @@ export const ChatForm: React.FC<ChatFormProps> = ({
                 data-testid="chat-form-textarea"
                 required={true}
                 // disabled={isStreaming}
-                disabled={limitReached}
                 {...props}
                 autoFocus={autoFocus}
                 style={{ boxShadow: "none", outline: "none" }}
diff --git a/refact-agent/gui/src/components/ChatForm/SuggestNewChat/SuggestNewChat.tsx b/refact-agent/gui/src/components/ChatForm/SuggestNewChat/SuggestNewChat.tsx
index 72776e0c4..359f78de6 100644
--- a/refact-agent/gui/src/components/ChatForm/SuggestNewChat/SuggestNewChat.tsx
+++ b/refact-agent/gui/src/components/ChatForm/SuggestNewChat/SuggestNewChat.tsx
@@ -8,13 +8,13 @@ import {
   useAppDispatch,
   useAppSelector,
   useCompressChat,
+  useLastSentCompressionStop,
 } from "../../../hooks";
 import { popBackTo, push } from "../../../features/Pages/pagesSlice";
 import { telemetryApi } from "../../../services/refact";
 import {
   newChatAction,
   selectChatId,
-  selectLastSentCompression,
   setIsNewChatSuggestionRejected,
 } from "../../../features/Chat";
 
@@ -40,8 +40,8 @@ export const SuggestNewChat = ({
   const [isRendered, setIsRendered] = useState(shouldBeVisible);
   const [isAnimating, setIsAnimating] = useState(false);
   const { compressChat, compressChatRequest } = useCompressChat();
+  const lastSentCompression = useLastSentCompressionStop();
 
-  const lastSentCompression = useAppSelector(selectLastSentCompression);
   useEffect(() => {
     if (shouldBeVisible) {
       setIsRendered(true);
@@ -64,6 +64,7 @@ export const SuggestNewChat = ({
 
   const handleClose = () => {
     dispatch(setIsNewChatSuggestionRejected({ chatId, value: true }));
+
     void sendTelemetryEvent({
       scope: `dismissedNewChatSuggestionWarning`,
       success: true,
@@ -118,27 +119,28 @@ export const SuggestNewChat = ({
           <Link size="1" onClick={onCreateNewChat} color="indigo">
             Start a new chat
           </Link>
-          {lastSentCompression && lastSentCompression !== "absent" && (
-            <Link
-              size="1"
-              onClick={() => {
-                if (compressChatRequest.isLoading) return;
-                void compressChat();
-              }}
-              color="indigo"
-              asChild
-            >
-              <Flex
-                align="center"
-                justify="start"
-                gap="1"
-                display="inline-flex"
+          {lastSentCompression.strength &&
+            lastSentCompression.strength !== "absent" && (
+              <Link
+                size="1"
+                onClick={() => {
+                  if (compressChatRequest.isLoading) return;
+                  void compressChat();
+                }}
+                color="indigo"
+                asChild
               >
-                <ArchiveIcon style={{ alignSelf: "start" }} />
-                Compress and open in a new chat.
-              </Flex>
-            </Link>
-          )}
+                <Flex
+                  align="center"
+                  justify="start"
+                  gap="1"
+                  display="inline-flex"
+                >
+                  <ArchiveIcon style={{ alignSelf: "start" }} />
+                  Compress and open in a new chat.
+                </Flex>
+              </Link>
+            )}
         </Flex>
         <Box position="absolute" top="1" right="1">
           <IconButton
diff --git a/refact-agent/gui/src/components/UsageCounter/useUsageCounter.ts b/refact-agent/gui/src/components/UsageCounter/useUsageCounter.ts
index e322ed93a..9d2c4ec6a 100644
--- a/refact-agent/gui/src/components/UsageCounter/useUsageCounter.ts
+++ b/refact-agent/gui/src/components/UsageCounter/useUsageCounter.ts
@@ -3,9 +3,9 @@ import {
   selectIsStreaming,
   selectIsWaiting,
   selectMessages,
-  selectLastSentCompression,
+  // selectLastSentCompression,
 } from "../../features/Chat";
-import { useAppSelector } from "../../hooks";
+import { useAppSelector, useLastSentCompressionStop } from "../../hooks";
 import {
   calculateUsageInputTokens,
   mergeUsages,
@@ -15,7 +15,7 @@ import { isAssistantMessage } from "../../services/refact";
 export function useUsageCounter() {
   const isStreaming = useAppSelector(selectIsStreaming);
   const isWaiting = useAppSelector(selectIsWaiting);
-  const lastSentCompression = useAppSelector(selectLastSentCompression);
+  const compressionStop = useLastSentCompressionStop();
   const messages = useAppSelector(selectMessages);
   const assistantMessages = messages.filter(isAssistantMessage);
   const usages = assistantMessages.map((msg) => msg.usage);
@@ -33,21 +33,23 @@ export function useUsageCounter() {
   }, [currentThreadUsage]);
 
   const isOverflown = useMemo(() => {
-    if (lastSentCompression === "low") return true;
-    if (lastSentCompression === "medium") return true;
-    if (lastSentCompression === "high") return true;
+    if (!compressionStop.stopped) return false;
+    if (compressionStop.strength === "low") return true;
+    if (compressionStop.strength === "medium") return true;
+    if (compressionStop.strength === "high") return true;
     return false;
-  }, [lastSentCompression]);
+  }, [compressionStop.stopped, compressionStop.strength]);
 
   const isWarning = useMemo(() => {
-    if (lastSentCompression === "medium") return true;
-    if (lastSentCompression === "high") return true;
+    if (!compressionStop.stopped) return false;
+    if (compressionStop.strength === "medium") return true;
+    if (compressionStop.strength === "high") return true;
     return false;
-  }, [lastSentCompression]);
+  }, [compressionStop.stopped, compressionStop.strength]);
 
   const shouldShow = useMemo(() => {
     return messages.length > 0 && !isStreaming && !isWaiting;
-  }, [messages, isStreaming, isWaiting]);
+  }, [messages.length, isStreaming, isWaiting]);
 
   return {
     shouldShow,
diff --git a/refact-agent/gui/src/features/Chat/Thread/actions.ts b/refact-agent/gui/src/features/Chat/Thread/actions.ts
index c10b0460f..bbc08eca7 100644
--- a/refact-agent/gui/src/features/Chat/Thread/actions.ts
+++ b/refact-agent/gui/src/features/Chat/Thread/actions.ts
@@ -63,6 +63,7 @@ export const setLastUserMessageId = createAction<PayloadWithChatAndMessageId>(
   "chatThread/setLastUserMessageId",
 );
 
+// TBD: only used when `/links` suggests a new chat.
 export const setIsNewChatSuggested = createAction<PayloadWithChatAndBoolean>(
   "chatThread/setIsNewChatSuggested",
 );
@@ -162,6 +163,10 @@ export const setIncreaseMaxTokens = createAction<boolean>(
   "chatThread/setIncreaseMaxTokens",
 );
 
+export const setThreadPaused = createAction<boolean>(
+  "chatThread/setThreadPaused",
+);
+
 // TODO: This is the circular dep when imported from hooks :/
 const createAppAsyncThunk = createAsyncThunk.withTypes<{
   state: RootState;
diff --git a/refact-agent/gui/src/features/Chat/Thread/reducer.ts b/refact-agent/gui/src/features/Chat/Thread/reducer.ts
index 87066d125..d0692fd28 100644
--- a/refact-agent/gui/src/features/Chat/Thread/reducer.ts
+++ b/refact-agent/gui/src/features/Chat/Thread/reducer.ts
@@ -39,6 +39,7 @@ import {
   setIsNewChatSuggestionRejected,
   upsertToolCall,
   setIncreaseMaxTokens,
+  setThreadPaused,
 } from "./actions";
 import { formatChatResponse } from "./utils";
 import {
@@ -78,6 +79,7 @@ const createChatThread = (
     boost_reasoning: false,
     automatic_patch: false,
     increase_max_tokens: false,
+    paused: false,
   };
   return chat;
 };
@@ -241,6 +243,7 @@ export const chatReducer = createReducer(initialState, (builder) => {
 
   builder.addCase(setIsNewChatSuggested, (state, action) => {
     if (state.thread.id !== action.payload.chatId) return state;
+    state.thread.paused = true;
     state.thread.new_chat_suggested = {
       wasSuggested: action.payload.value,
     };
@@ -248,6 +251,7 @@ export const chatReducer = createReducer(initialState, (builder) => {
 
   builder.addCase(setIsNewChatSuggestionRejected, (state, action) => {
     if (state.thread.id !== action.payload.chatId) return state;
+    state.thread.paused = false;
     state.thread.new_chat_suggested = {
       ...state.thread.new_chat_suggested,
       wasRejectedByUser: action.payload.value,
@@ -275,6 +279,7 @@ export const chatReducer = createReducer(initialState, (builder) => {
     state.streaming = true;
     state.thread.read = false;
     state.prevent_send = false;
+    state.thread.paused = false;
   });
 
   builder.addCase(removeChatFromCache, (state, action) => {
@@ -422,6 +427,10 @@ export const chatReducer = createReducer(initialState, (builder) => {
     state.thread.increase_max_tokens = action.payload;
   });
 
+  builder.addCase(setThreadPaused, (state, action) => {
+    state.thread.paused = action.payload;
+  });
+
   builder.addMatcher(
     capsApi.endpoints.getCaps.matchFulfilled,
     (state, action) => {
diff --git a/refact-agent/gui/src/features/Chat/Thread/selectors.ts b/refact-agent/gui/src/features/Chat/Thread/selectors.ts
index 56393d114..563ba6102 100644
--- a/refact-agent/gui/src/features/Chat/Thread/selectors.ts
+++ b/refact-agent/gui/src/features/Chat/Thread/selectors.ts
@@ -1,9 +1,9 @@
 import { RootState } from "../../../app/store";
 import { createSelector } from "@reduxjs/toolkit";
 import {
+  CompressionStrength,
   isToolMessage,
   isUserMessage,
-  UserMessage,
 } from "../../../services/refact/types";
 
 export const selectThread = (state: RootState) => state.chat.thread;
@@ -23,6 +23,7 @@ export const selectCheckpointsEnabled = (state: RootState) =>
 export const selectThreadBoostReasoning = (state: RootState) =>
   state.chat.thread.boost_reasoning;
 
+// TBD: only used when `/links` suggests a new chat.
 export const selectThreadNewChatSuggested = (state: RootState) =>
   state.chat.thread.new_chat_suggested;
 export const selectThreadMaximumTokens = (state: RootState) =>
@@ -77,14 +78,24 @@ export const selectThreadMode = createSelector(
 export const selectLastSentCompression = createSelector(
   selectMessages,
   (messages) => {
-    const lastUserMessage = messages.reduce<null | UserMessage>(
+    const lastCompression = messages.reduce<null | CompressionStrength>(
       (acc, message) => {
-        if (isUserMessage(message)) return message;
+        if (isUserMessage(message) && message.compression_strength) {
+          return message.compression_strength;
+        }
+        if (isToolMessage(message) && message.content.compression_strength) {
+          return message.content.compression_strength;
+        }
         return acc;
       },
       null,
     );
 
-    return lastUserMessage?.compression_strength ?? null;
+    return lastCompression;
   },
 );
+
+export const selectThreadPaused = createSelector(
+  selectThread,
+  (thread) => thread.paused ?? false,
+);
diff --git a/refact-agent/gui/src/features/Chat/Thread/types.ts b/refact-agent/gui/src/features/Chat/Thread/types.ts
index 6a038aacf..7c907148b 100644
--- a/refact-agent/gui/src/features/Chat/Thread/types.ts
+++ b/refact-agent/gui/src/features/Chat/Thread/types.ts
@@ -29,6 +29,7 @@ export type ChatThread = {
   currentMaximumContextTokens?: number;
   currentMessageContextTokens?: number;
   increase_max_tokens?: boolean;
+  paused?: boolean;
 };
 
 export type SuggestedChat = {
diff --git a/refact-agent/gui/src/features/Chat/Thread/utils.ts b/refact-agent/gui/src/features/Chat/Thread/utils.ts
index f6a741f47..520aa3321 100644
--- a/refact-agent/gui/src/features/Chat/Thread/utils.ts
+++ b/refact-agent/gui/src/features/Chat/Thread/utils.ts
@@ -182,7 +182,8 @@ export function formatChatResponse(
   }
 
   if (isToolResponse(response)) {
-    const { tool_call_id, content, finish_reason } = response;
+    const { tool_call_id, content, finish_reason, compression_strength } =
+      response;
     const filteredMessages = finishToolCallInMessages(messages, tool_call_id);
     const toolResult: ToolResult =
       typeof content === "string"
@@ -190,11 +191,13 @@ export function formatChatResponse(
             tool_call_id,
             content,
             finish_reason,
+            compression_strength,
           }
         : {
             tool_call_id,
             content,
             finish_reason,
+            compression_strength,
           };
 
     return [...filteredMessages, { role: response.role, content: toolResult }];
diff --git a/refact-agent/gui/src/hooks/index.ts b/refact-agent/gui/src/hooks/index.ts
index 3ffb2015a..6dfb21e34 100644
--- a/refact-agent/gui/src/hooks/index.ts
+++ b/refact-agent/gui/src/hooks/index.ts
@@ -34,4 +34,4 @@ export * from "./useResizeObserver";
 export * from "./useCompressChat";
 export * from "./useAutoFocusOnce";
 export * from "./useHideScroll";
-export * from "./useTotalTokenUsage";
+export * from "./useCompressionStop";
diff --git a/refact-agent/gui/src/hooks/useCompressionStop.ts b/refact-agent/gui/src/hooks/useCompressionStop.ts
new file mode 100644
index 000000000..aad6687be
--- /dev/null
+++ b/refact-agent/gui/src/hooks/useCompressionStop.ts
@@ -0,0 +1,45 @@
+import { useEffect, useCallback, useMemo } from "react";
+import { useAppSelector } from "./useAppSelector";
+import { useAppDispatch } from "./useAppDispatch";
+import {
+  selectLastSentCompression,
+  selectMessages,
+  selectThreadPaused,
+  setThreadPaused,
+} from "../features/Chat";
+import { takeFromEndWhile } from "../utils";
+import { isUserMessage } from "../events";
+
+export function useLastSentCompressionStop() {
+  const dispatch = useAppDispatch();
+  const lastSentCompression = useAppSelector(selectLastSentCompression);
+  const messages = useAppSelector(selectMessages);
+  const stopped = useAppSelector(selectThreadPaused);
+  useEffect(() => {
+    if (lastSentCompression && lastSentCompression !== "absent" && !stopped) {
+      dispatch(setThreadPaused(true));
+    }
+  }, [dispatch, lastSentCompression, stopped]);
+
+  const messagesFromLastUserMessage = useMemo(() => {
+    return takeFromEndWhile(messages, (message) => !isUserMessage(message))
+      .length;
+  }, [messages]);
+
+  useEffect(() => {
+    if (
+      lastSentCompression &&
+      lastSentCompression !== "absent" &&
+      messagesFromLastUserMessage >= 40 &&
+      !stopped
+    ) {
+      dispatch(setThreadPaused(true));
+    }
+  }, [dispatch, lastSentCompression, messagesFromLastUserMessage, stopped]);
+
+  const resume = useCallback(() => {
+    dispatch(setThreadPaused(false));
+  }, [dispatch]);
+
+  return { stopped, resume, strength: lastSentCompression };
+}
diff --git a/refact-agent/gui/src/hooks/useSendChatRequest.ts b/refact-agent/gui/src/hooks/useSendChatRequest.ts
index c515ddd4f..05d74e4c4 100644
--- a/refact-agent/gui/src/hooks/useSendChatRequest.ts
+++ b/refact-agent/gui/src/hooks/useSendChatRequest.ts
@@ -56,7 +56,7 @@ import {
 
 import { v4 as uuidv4 } from "uuid";
 import { upsertToolCallIntoHistory } from "../features/History/historySlice";
-import { useTotalTokenUsage } from "./useTotalTokenUsage";
+import { useLastSentCompressionStop } from "./useCompressionStop";
 
 type SubmitHandlerParams =
   | {
@@ -356,7 +356,7 @@ export function useAutoSend() {
   const sendImmediately = useAppSelector(selectSendImmediately);
   const wasInteracted = useAppSelector(getToolsInteractionStatus); // shows if tool confirmation popup was interacted by user
   const areToolsConfirmed = useAppSelector(getToolsConfirmationStatus);
-  const { limitReached } = useTotalTokenUsage();
+  const compressionStop = useLastSentCompressionStop();
   const { sendMessages, abort, messagesWithSystemPrompt } =
     useSendChatRequest();
   // TODO: make a selector for this, or show tool formation
@@ -381,7 +381,7 @@ export function useAutoSend() {
       const lastMessage = currentMessages.slice(-1)[0];
       // here ish
       if (
-        !limitReached &&
+        !compressionStop.stopped &&
         isAssistantMessage(lastMessage) &&
         lastMessage.tool_calls &&
         lastMessage.tool_calls.length > 0
@@ -418,6 +418,6 @@ export function useAutoSend() {
     isIntegration,
     thread.mode,
     thread,
-    limitReached,
+    compressionStop.stopped,
   ]);
 }
diff --git a/refact-agent/gui/src/hooks/useTotalTokenUsage.ts b/refact-agent/gui/src/hooks/useTotalTokenUsage.ts
deleted file mode 100644
index 5c2a3f7e2..000000000
--- a/refact-agent/gui/src/hooks/useTotalTokenUsage.ts
+++ /dev/null
@@ -1,47 +0,0 @@
-import { useMemo } from "react";
-import { useAppSelector } from "./useAppSelector";
-import { isAssistantMessage } from "../events";
-import { selectMessages } from "../features/Chat";
-import { type Usage } from "../services/refact/chat";
-import {
-  calculateUsageInputTokens,
-  mergeUsages,
-} from "../utils/calculateUsageInputTokens";
-
-const TOKEN_LIMIT = 200_000;
-export function useTotalTokenUsage() {
-  const messages = useAppSelector(selectMessages);
-
-  const summedUsages = useMemo(() => {
-    const usages = messages.reduce<Usage[]>((acc, message) => {
-      if (isAssistantMessage(message) && message.usage) {
-        return [...acc, message.usage];
-      }
-      return acc;
-    }, []);
-    return mergeUsages(usages);
-  }, [messages]);
-
-  const tokens = useMemo(() => {
-    if (!summedUsages) return 0;
-    return calculateUsageInputTokens({
-      keys: [
-        "prompt_tokens",
-        "cache_creation_input_tokens",
-        "cache_read_input_tokens",
-      ],
-      usage: summedUsages,
-    });
-  }, [summedUsages]);
-
-  const limitReached = useMemo(() => {
-    return tokens >= TOKEN_LIMIT;
-  }, [tokens]);
-
-  return {
-    summedUsages,
-    tokens,
-    limitReached,
-    limit: TOKEN_LIMIT,
-  };
-}
diff --git a/refact-agent/gui/src/services/refact/types.ts b/refact-agent/gui/src/services/refact/types.ts
index 0fafc5b1a..eef22c1a4 100644
--- a/refact-agent/gui/src/services/refact/types.ts
+++ b/refact-agent/gui/src/services/refact/types.ts
@@ -70,6 +70,7 @@ export interface BaseToolResult {
   tool_call_id: string;
   finish_reason?: string; // "call_failed" | "call_worked";
   content: ToolContent;
+  compression_strength?: CompressionStrength;
 }
 
 export interface SingleModelToolResult extends BaseToolResult {