diff --git a/.changeset/large-bags-send.md b/.changeset/large-bags-send.md
new file mode 100644
index 00000000000..93e49b2388e
--- /dev/null
+++ b/.changeset/large-bags-send.md
@@ -0,0 +1,5 @@
+---
+"roo-cline": patch
+---
+
+Adds experimental feature to intelligently condense the task context
diff --git a/evals/packages/types/src/roo-code.ts b/evals/packages/types/src/roo-code.ts
index cef4056ce73..0e1675b1dec 100644
--- a/evals/packages/types/src/roo-code.ts
+++ b/evals/packages/types/src/roo-code.ts
@@ -297,7 +297,7 @@ export type CommandExecutionStatus = z.infer<typeof commandExecutionStatusSchema
  * ExperimentId
  */
 
-export const experimentIds = ["powerSteering"] as const
+export const experimentIds = ["autoCondenseContext", "powerSteering"] as const
 
 export const experimentIdsSchema = z.enum(experimentIds)
 
@@ -308,6 +308,7 @@ export type ExperimentId = z.infer<typeof experimentIdsSchema>
  */
 
 const experimentsSchema = z.object({
+	autoCondenseContext: z.boolean(),
 	powerSteering: z.boolean(),
 })
 
diff --git a/src/api/transform/__tests__/image-cleaning.test.ts b/src/api/transform/__tests__/image-cleaning.test.ts
new file mode 100644
index 00000000000..cbb318531a1
--- /dev/null
+++ b/src/api/transform/__tests__/image-cleaning.test.ts
@@ -0,0 +1,336 @@
+import { ApiHandler } from "../.."
+import { ApiMessage } from "../../../core/task-persistence/apiMessages"
+import { maybeRemoveImageBlocks } from "../image-cleaning"
+import { ModelInfo } from "../../../shared/api"
+
+describe("maybeRemoveImageBlocks", () => {
+	// Mock ApiHandler factory function
+	const createMockApiHandler = (supportsImages: boolean): ApiHandler => {
+		return {
+			getModel: jest.fn().mockReturnValue({
+				id: "test-model",
+				info: {
+					supportsImages,
+				} as ModelInfo,
+			}),
+			createMessage: jest.fn(),
+			countTokens: jest.fn(),
+		}
+	}
+
+	it("should handle empty messages array", () => {
+		const apiHandler = createMockApiHandler(true)
+		const messages: ApiMessage[] = []
+
+		const result = maybeRemoveImageBlocks(messages, apiHandler)
+
+		expect(result).toEqual([])
+		// No need to check if getModel was called since there are no messages to process
+	})
+
+	it("should not modify messages with no image blocks", () => {
+		const apiHandler = createMockApiHandler(true)
+		const messages: ApiMessage[] = [
+			{
+				role: "user",
+				content: "Hello, world!",
+			},
+			{
+				role: "assistant",
+				content: "Hi there!",
+			},
+		]
+
+		const result = maybeRemoveImageBlocks(messages, apiHandler)
+
+		expect(result).toEqual(messages)
+		// getModel is only called when content is an array, which is not the case here
+	})
+
+	it("should not modify messages with array content but no image blocks", () => {
+		const apiHandler = createMockApiHandler(true)
+		const messages: ApiMessage[] = [
+			{
+				role: "user",
+				content: [
+					{
+						type: "text",
+						text: "Hello, world!",
+					},
+					{
+						type: "text",
+						text: "How are you?",
+					},
+				],
+			},
+		]
+
+		const result = maybeRemoveImageBlocks(messages, apiHandler)
+
+		expect(result).toEqual(messages)
+		expect(apiHandler.getModel).toHaveBeenCalled()
+	})
+
+	it("should not modify image blocks when API handler supports images", () => {
+		const apiHandler = createMockApiHandler(true)
+		const messages: ApiMessage[] = [
+			{
+				role: "user",
+				content: [
+					{
+						type: "text",
+						text: "Check out this image:",
+					},
+					{
+						type: "image",
+						source: {
+							type: "base64",
+							media_type: "image/jpeg",
+							data: "base64-encoded-image-data",
+						},
+					},
+				],
+			},
+		]
+
+		const result = maybeRemoveImageBlocks(messages, apiHandler)
+
+		// Should not modify the messages since the API handler supports images
+		expect(result).toEqual(messages)
+		expect(apiHandler.getModel).toHaveBeenCalled()
+	})
+
+	it("should convert image blocks to text descriptions when API handler doesn't support images", () => {
+		const apiHandler = createMockApiHandler(false)
+		const messages: ApiMessage[] = [
+			{
+				role: "user",
+				content: [
+					{
+						type: "text",
+						text: "Check out this image:",
+					},
+					{
+						type: "image",
+						source: {
+							type: "base64",
+							media_type: "image/jpeg",
+							data: "base64-encoded-image-data",
+						},
+					},
+				],
+			},
+		]
+
+		const result = maybeRemoveImageBlocks(messages, apiHandler)
+
+		// Should convert image blocks to text descriptions
+		expect(result).toEqual([
+			{
+				role: "user",
+				content: [
+					{
+						type: "text",
+						text: "Check out this image:",
+					},
+					{
+						type: "text",
+						text: "[Referenced image in conversation]",
+					},
+				],
+			},
+		])
+		expect(apiHandler.getModel).toHaveBeenCalled()
+	})
+
+	it("should handle mixed content messages with multiple text and image blocks", () => {
+		const apiHandler = createMockApiHandler(false)
+		const messages: ApiMessage[] = [
+			{
+				role: "user",
+				content: [
+					{
+						type: "text",
+						text: "Here are some images:",
+					},
+					{
+						type: "image",
+						source: {
+							type: "base64",
+							media_type: "image/jpeg",
+							data: "image-data-1",
+						},
+					},
+					{
+						type: "text",
+						text: "And another one:",
+					},
+					{
+						type: "image",
+						source: {
+							type: "base64",
+							media_type: "image/png",
+							data: "image-data-2",
+						},
+					},
+				],
+			},
+		]
+
+		const result = maybeRemoveImageBlocks(messages, apiHandler)
+
+		// Should convert all image blocks to text descriptions
+		expect(result).toEqual([
+			{
+				role: "user",
+				content: [
+					{
+						type: "text",
+						text: "Here are some images:",
+					},
+					{
+						type: "text",
+						text: "[Referenced image in conversation]",
+					},
+					{
+						type: "text",
+						text: "And another one:",
+					},
+					{
+						type: "text",
+						text: "[Referenced image in conversation]",
+					},
+				],
+			},
+		])
+		expect(apiHandler.getModel).toHaveBeenCalled()
+	})
+
+	it("should handle multiple messages with image blocks", () => {
+		const apiHandler = createMockApiHandler(false)
+		const messages: ApiMessage[] = [
+			{
+				role: "user",
+				content: [
+					{
+						type: "text",
+						text: "Here's an image:",
+					},
+					{
+						type: "image",
+						source: {
+							type: "base64",
+							media_type: "image/jpeg",
+							data: "image-data-1",
+						},
+					},
+				],
+			},
+			{
+				role: "assistant",
+				content: "I see the image!",
+			},
+			{
+				role: "user",
+				content: [
+					{
+						type: "text",
+						text: "Here's another image:",
+					},
+					{
+						type: "image",
+						source: {
+							type: "base64",
+							media_type: "image/png",
+							data: "image-data-2",
+						},
+					},
+				],
+			},
+		]
+
+		const result = maybeRemoveImageBlocks(messages, apiHandler)
+
+		// Should convert all image blocks to text descriptions
+		expect(result).toEqual([
+			{
+				role: "user",
+				content: [
+					{
+						type: "text",
+						text: "Here's an image:",
+					},
+					{
+						type: "text",
+						text: "[Referenced image in conversation]",
+					},
+				],
+			},
+			{
+				role: "assistant",
+				content: "I see the image!",
+			},
+			{
+				role: "user",
+				content: [
+					{
+						type: "text",
+						text: "Here's another image:",
+					},
+					{
+						type: "text",
+						text: "[Referenced image in conversation]",
+					},
+				],
+			},
+		])
+		expect(apiHandler.getModel).toHaveBeenCalled()
+	})
+
+	it("should preserve additional message properties", () => {
+		const apiHandler = createMockApiHandler(false)
+		const messages: ApiMessage[] = [
+			{
+				role: "user",
+				content: [
+					{
+						type: "text",
+						text: "Here's an image:",
+					},
+					{
+						type: "image",
+						source: {
+							type: "base64",
+							media_type: "image/jpeg",
+							data: "image-data",
+						},
+					},
+				],
+				ts: 1620000000000,
+				isSummary: true,
+			},
+		]
+
+		const result = maybeRemoveImageBlocks(messages, apiHandler)
+
+		// Should convert image blocks to text descriptions while preserving additional properties
+		expect(result).toEqual([
+			{
+				role: "user",
+				content: [
+					{
+						type: "text",
+						text: "Here's an image:",
+					},
+					{
+						type: "text",
+						text: "[Referenced image in conversation]",
+					},
+				],
+				ts: 1620000000000,
+				isSummary: true,
+			},
+		])
+		expect(apiHandler.getModel).toHaveBeenCalled()
+	})
+})
diff --git a/src/api/transform/image-cleaning.ts b/src/api/transform/image-cleaning.ts
new file mode 100644
index 00000000000..e5987bb59e5
--- /dev/null
+++ b/src/api/transform/image-cleaning.ts
@@ -0,0 +1,28 @@
+import { ApiHandler } from ".."
+import { ApiMessage } from "../../core/task-persistence/apiMessages"
+
+/* Removes image blocks from messages if they are not supported by the Api Handler */
+export function maybeRemoveImageBlocks(messages: ApiMessage[], apiHandler: ApiHandler): ApiMessage[] {
+	return messages.map((message) => {
+		// Handle array content (could contain image blocks).
+		let { content } = message
+		if (Array.isArray(content)) {
+			if (!apiHandler.getModel().info.supportsImages) {
+				// Convert image blocks to text descriptions.
+				content = content.map((block) => {
+					if (block.type === "image") {
+						// Convert image blocks to text descriptions.
+						// Note: We can't access the actual image content/url due to API limitations,
+						// but we can indicate that an image was present in the conversation.
+						return {
+							type: "text",
+							text: "[Referenced image in conversation]",
+						}
+					}
+					return block
+				})
+			}
+		}
+		return { ...message, content }
+	})
+}
diff --git a/src/core/condense/__tests__/index.test.ts b/src/core/condense/__tests__/index.test.ts
new file mode 100644
index 00000000000..10769589f86
--- /dev/null
+++ b/src/core/condense/__tests__/index.test.ts
@@ -0,0 +1,228 @@
+import { describe, expect, it, jest, beforeEach } from "@jest/globals"
+import { ApiHandler } from "../../../api"
+import { ApiMessage } from "../../task-persistence/apiMessages"
+import { maybeRemoveImageBlocks } from "../../../api/transform/image-cleaning"
+import { summarizeConversation, getMessagesSinceLastSummary, N_MESSAGES_TO_KEEP } from "../index"
+
+// Mock dependencies
+jest.mock("../../../api/transform/image-cleaning", () => ({
+	maybeRemoveImageBlocks: jest.fn((messages: ApiMessage[], _apiHandler: ApiHandler) => [...messages]),
+}))
+
+describe("getMessagesSinceLastSummary", () => {
+	it("should return all messages when there is no summary", () => {
+		const messages: ApiMessage[] = [
+			{ role: "user", content: "Hello", ts: 1 },
+			{ role: "assistant", content: "Hi there", ts: 2 },
+			{ role: "user", content: "How are you?", ts: 3 },
+		]
+
+		const result = getMessagesSinceLastSummary(messages)
+		expect(result).toEqual(messages)
+	})
+
+	it("should return messages since the last summary", () => {
+		const messages: ApiMessage[] = [
+			{ role: "user", content: "Hello", ts: 1 },
+			{ role: "assistant", content: "Hi there", ts: 2 },
+			{ role: "assistant", content: "Summary of conversation", ts: 3, isSummary: true },
+			{ role: "user", content: "How are you?", ts: 4 },
+			{ role: "assistant", content: "I'm good", ts: 5 },
+		]
+
+		const result = getMessagesSinceLastSummary(messages)
+		expect(result).toEqual([
+			{ role: "assistant", content: "Summary of conversation", ts: 3, isSummary: true },
+			{ role: "user", content: "How are you?", ts: 4 },
+			{ role: "assistant", content: "I'm good", ts: 5 },
+		])
+	})
+
+	it("should handle multiple summary messages and return since the last one", () => {
+		const messages: ApiMessage[] = [
+			{ role: "user", content: "Hello", ts: 1 },
+			{ role: "assistant", content: "First summary", ts: 2, isSummary: true },
+			{ role: "user", content: "How are you?", ts: 3 },
+			{ role: "assistant", content: "Second summary", ts: 4, isSummary: true },
+			{ role: "user", content: "What's new?", ts: 5 },
+		]
+
+		const result = getMessagesSinceLastSummary(messages)
+		expect(result).toEqual([
+			{ role: "assistant", content: "Second summary", ts: 4, isSummary: true },
+			{ role: "user", content: "What's new?", ts: 5 },
+		])
+	})
+
+	it("should handle empty messages array", () => {
+		const result = getMessagesSinceLastSummary([])
+		expect(result).toEqual([])
+	})
+})
+
+describe("summarizeConversation", () => {
+	// Mock ApiHandler
+	let mockApiHandler: ApiHandler
+	let mockStream: AsyncGenerator<any, void, unknown>
+
+	beforeEach(() => {
+		// Reset mocks
+		jest.clearAllMocks()
+
+		// Setup mock stream
+		mockStream = (async function* () {
+			yield { type: "text" as const, text: "This is " }
+			yield { type: "text" as const, text: "a summary" }
+		})()
+
+		// Setup mock API handler
+		mockApiHandler = {
+			createMessage: jest.fn().mockReturnValue(mockStream),
+			countTokens: jest.fn().mockImplementation(() => Promise.resolve(100)),
+			getModel: jest.fn().mockReturnValue({
+				id: "test-model",
+				info: {
+					contextWindow: 8000,
+					supportsImages: true,
+					supportsComputerUse: true,
+					supportsVision: true,
+					maxTokens: 4000,
+					supportsPromptCache: true,
+					maxCachePoints: 10,
+					minTokensPerCachePoint: 100,
+					cachableFields: ["system", "messages"],
+				},
+			}),
+		} as unknown as ApiHandler
+	})
+
+	it("should not summarize when there are not enough messages", async () => {
+		const messages: ApiMessage[] = [
+			{ role: "user", content: "Hello", ts: 1 },
+			{ role: "assistant", content: "Hi there", ts: 2 },
+		]
+
+		const result = await summarizeConversation(messages, mockApiHandler)
+		expect(result).toEqual(messages)
+		expect(mockApiHandler.createMessage).not.toHaveBeenCalled()
+	})
+
+	it("should not summarize when there was a recent summary", async () => {
+		const messages: ApiMessage[] = [
+			{ role: "user", content: "Hello", ts: 1 },
+			{ role: "assistant", content: "Hi there", ts: 2 },
+			{ role: "user", content: "How are you?", ts: 3 },
+			{ role: "assistant", content: "I'm good", ts: 4 },
+			{ role: "user", content: "What's new?", ts: 5 },
+			{ role: "assistant", content: "Not much", ts: 6, isSummary: true }, // Recent summary
+			{ role: "user", content: "Tell me more", ts: 7 },
+		]
+
+		const result = await summarizeConversation(messages, mockApiHandler)
+		expect(result).toEqual(messages)
+		expect(mockApiHandler.createMessage).not.toHaveBeenCalled()
+	})
+
+	it("should summarize conversation and insert summary message", async () => {
+		const messages: ApiMessage[] = [
+			{ role: "user", content: "Hello", ts: 1 },
+			{ role: "assistant", content: "Hi there", ts: 2 },
+			{ role: "user", content: "How are you?", ts: 3 },
+			{ role: "assistant", content: "I'm good", ts: 4 },
+			{ role: "user", content: "What's new?", ts: 5 },
+			{ role: "assistant", content: "Not much", ts: 6 },
+			{ role: "user", content: "Tell me more", ts: 7 },
+		]
+
+		const result = await summarizeConversation(messages, mockApiHandler)
+
+		// Check that the API was called correctly
+		expect(mockApiHandler.createMessage).toHaveBeenCalled()
+		expect(maybeRemoveImageBlocks).toHaveBeenCalled()
+
+		// Verify the structure of the result
+		// The result should be: original messages (except last N) + summary + last N messages
+		expect(result.length).toBe(messages.length + 1) // Original + summary
+
+		// Check that the summary message was inserted correctly
+		const summaryMessage = result[result.length - N_MESSAGES_TO_KEEP - 1]
+		expect(summaryMessage.role).toBe("assistant")
+		expect(summaryMessage.content).toBe("This is a summary")
+		expect(summaryMessage.isSummary).toBe(true)
+
+		// Check that the last N_MESSAGES_TO_KEEP messages are preserved
+		const lastMessages = messages.slice(-N_MESSAGES_TO_KEEP)
+		expect(result.slice(-N_MESSAGES_TO_KEEP)).toEqual(lastMessages)
+	})
+
+	it("should handle empty summary response", async () => {
+		// We need enough messages to trigger summarization
+		const messages: ApiMessage[] = [
+			{ role: "user", content: "Hello", ts: 1 },
+			{ role: "assistant", content: "Hi there", ts: 2 },
+			{ role: "user", content: "How are you?", ts: 3 },
+			{ role: "assistant", content: "I'm good", ts: 4 },
+			{ role: "user", content: "What's new?", ts: 5 },
+			{ role: "assistant", content: "Not much", ts: 6 },
+			{ role: "user", content: "Tell me more", ts: 7 },
+		]
+
+		// Mock console.warn before we call the function
+		const originalWarn = console.warn
+		const mockWarn = jest.fn()
+		console.warn = mockWarn
+
+		// Setup empty summary response
+		const emptyStream = (async function* () {
+			yield { type: "text" as const, text: "" }
+		})()
+
+		// Create a new mock for createMessage that returns empty stream
+		const createMessageMock = jest.fn().mockReturnValue(emptyStream)
+		mockApiHandler.createMessage = createMessageMock as any
+
+		// We need to mock maybeRemoveImageBlocks to return the expected messages
+		;(maybeRemoveImageBlocks as jest.Mock).mockImplementationOnce((messages: any) => {
+			return messages.map(({ role, content }: { role: string; content: any }) => ({ role, content }))
+		})
+
+		const result = await summarizeConversation(messages, mockApiHandler)
+
+		// Should return original messages when summary is empty
+		expect(result).toEqual(messages)
+		expect(mockWarn).toHaveBeenCalledWith("Received empty summary from API")
+
+		// Restore console.warn
+		console.warn = originalWarn
+	})
+
+	it("should correctly format the request to the API", async () => {
+		const messages: ApiMessage[] = [
+			{ role: "user", content: "Hello", ts: 1 },
+			{ role: "assistant", content: "Hi there", ts: 2 },
+			{ role: "user", content: "How are you?", ts: 3 },
+			{ role: "assistant", content: "I'm good", ts: 4 },
+			{ role: "user", content: "What's new?", ts: 5 },
+			{ role: "assistant", content: "Not much", ts: 6 },
+			{ role: "user", content: "Tell me more", ts: 7 },
+		]
+
+		await summarizeConversation(messages, mockApiHandler)
+
+		// Verify the final request message
+		const expectedFinalMessage = {
+			role: "user",
+			content: "Summarize the conversation so far, as described in the prompt instructions.",
+		}
+
+		// Verify that createMessage was called with the correct prompt
+		expect(mockApiHandler.createMessage).toHaveBeenCalledWith(
+			expect.stringContaining("Your task is to create a detailed summary of the conversation"),
+			expect.any(Array),
+		)
+
+		// Check that maybeRemoveImageBlocks was called with the correct messages
+		const mockCallArgs = (maybeRemoveImageBlocks as jest.Mock).mock.calls[0][0] as any[]
+		expect(mockCallArgs[mockCallArgs.length - 1]).toEqual(expectedFinalMessage)
+	})
+})
diff --git a/src/core/condense/index.ts b/src/core/condense/index.ts
new file mode 100644
index 00000000000..2a88dbfccee
--- /dev/null
+++ b/src/core/condense/index.ts
@@ -0,0 +1,105 @@
+import Anthropic from "@anthropic-ai/sdk"
+import { ApiHandler } from "../../api"
+import { ApiMessage } from "../task-persistence/apiMessages"
+import { maybeRemoveImageBlocks } from "../../api/transform/image-cleaning"
+
+export const N_MESSAGES_TO_KEEP = 3
+
+const SUMMARY_PROMPT = `\
+Your task is to create a detailed summary of the conversation so far, paying close attention to the user's explicit requests and your previous actions.
+This summary should be thorough in capturing technical details, code patterns, and architectural decisions that would be essential for continuing with the conversation and supporting any continuing tasks.
+
+Your summary should be structured as follows:
+Context: The context to continue the conversation with. If applicable based on the current task, this should include:
+  1. Previous Conversation: High level details about what was discussed throughout the entire conversation with the user. This should be written to allow someone to be able to follow the general overarching conversation flow.
+  2. Current Work: Describe in detail what was being worked on prior to this request to summarize the conversation. Pay special attention to the more recent messages in the conversation.
+  3. Key Technical Concepts: List all important technical concepts, technologies, coding conventions, and frameworks discussed, which might be relevant for continuing with this work.
+  4. Relevant Files and Code: If applicable, enumerate specific files and code sections examined, modified, or created for the task continuation. Pay special attention to the most recent messages and changes.
+  5. Problem Solving: Document problems solved thus far and any ongoing troubleshooting efforts.
+  6. Pending Tasks and Next Steps: Outline all pending tasks that you have explicitly been asked to work on, as well as list the next steps you will take for all outstanding work, if applicable. Include code snippets where they add clarity. For any next steps, include direct quotes from the most recent conversation showing exactly what task you were working on and where you left off. This should be verbatim to ensure there's no information loss in context between tasks.
+
+Example summary structure:
+1. Previous Conversation:
+  [Detailed description]
+2. Current Work:
+  [Detailed description]
+3. Key Technical Concepts:
+  - [Concept 1]
+  - [Concept 2]
+  - [...]
+4. Relevant Files and Code:
+  - [File Name 1]
+    - [Summary of why this file is important]
+    - [Summary of the changes made to this file, if any]
+    - [Important Code Snippet]
+  - [File Name 2]
+    - [Important Code Snippet]
+  - [...]
+5. Problem Solving:
+  [Detailed description]
+6. Pending Tasks and Next Steps:
+  - [Task 1 details & next steps]
+  - [Task 2 details & next steps]
+  - [...]
+
+Output only the summary of the conversation so far, without any additional commentary or explanation.
+`
+
+/**
+ * Summarizes the conversation messages using an LLM call
+ *
+ * @param {ApiMessage[]} messages - The conversation messages
+ * @param {ApiHandler} apiHandler - The API handler to use for token counting.
+ * @returns {ApiMessage[]} - The input messages, potentially including a new summary message before the last message.
+ */
+export async function summarizeConversation(messages: ApiMessage[], apiHandler: ApiHandler): Promise<ApiMessage[]> {
+	const messagesToSummarize = getMessagesSinceLastSummary(messages.slice(0, -N_MESSAGES_TO_KEEP))
+	if (messagesToSummarize.length <= 1) {
+		return messages // Not enough messages to warrant a summary
+	}
+	const keepMessages = messages.slice(-N_MESSAGES_TO_KEEP)
+	for (const message of keepMessages) {
+		if (message.isSummary) {
+			return messages // We recently summarized these messages; it's too soon to summarize again.
+		}
+	}
+	const finalRequestMessage: Anthropic.MessageParam = {
+		role: "user",
+		content: "Summarize the conversation so far, as described in the prompt instructions.",
+	}
+	const requestMessages = maybeRemoveImageBlocks([...messagesToSummarize, finalRequestMessage], apiHandler).map(
+		({ role, content }) => ({ role, content }),
+	)
+	// Note: this doesn't need to be a stream, consider using something like apiHandler.completePrompt
+	const stream = apiHandler.createMessage(SUMMARY_PROMPT, requestMessages)
+	let summary = ""
+	// TODO(canyon): compute usage and cost for this operation and update the global metrics.
+	for await (const chunk of stream) {
+		if (chunk.type === "text") {
+			summary += chunk.text
+		}
+	}
+	summary = summary.trim()
+	if (summary.length === 0) {
+		console.warn("Received empty summary from API")
+		return messages
+	}
+	const summaryMessage: ApiMessage = {
+		role: "assistant",
+		content: summary,
+		ts: keepMessages[0].ts,
+		isSummary: true,
+	}
+
+	return [...messages.slice(0, -N_MESSAGES_TO_KEEP), summaryMessage, ...keepMessages]
+}
+
+/* Returns the list of all messages since the last summary message, including the summary. Returns all messages if there is no summary. */
+export function getMessagesSinceLastSummary(messages: ApiMessage[]): ApiMessage[] {
+	let lastSummaryIndexReverse = [...messages].reverse().findIndex((message) => message.isSummary)
+	if (lastSummaryIndexReverse === -1) {
+		return messages
+	}
+	const lastSummaryIndex = messages.length - lastSummaryIndexReverse - 1
+	return messages.slice(lastSummaryIndex)
+}
diff --git a/src/core/sliding-window/__tests__/sliding-window.test.ts b/src/core/sliding-window/__tests__/sliding-window.test.ts
index 16af2d46301..7890b55ec88 100644
--- a/src/core/sliding-window/__tests__/sliding-window.test.ts
+++ b/src/core/sliding-window/__tests__/sliding-window.test.ts
@@ -10,6 +10,7 @@ import {
 	truncateConversation,
 	truncateConversationIfNeeded,
 } from "../index"
+import { ApiMessage } from "../../task-persistence/apiMessages"
 
 // Create a mock ApiHandler for testing
 class MockApiHandler extends BaseProvider {
@@ -41,7 +42,7 @@ const mockApiHandler = new MockApiHandler()
  */
 describe("truncateConversation", () => {
 	it("should retain the first message", () => {
-		const messages: Anthropic.Messages.MessageParam[] = [
+		const messages: ApiMessage[] = [
 			{ role: "user", content: "First message" },
 			{ role: "assistant", content: "Second message" },
 			{ role: "user", content: "Third message" },
@@ -58,7 +59,7 @@ describe("truncateConversation", () => {
 	})
 
 	it("should remove the specified fraction of messages (rounded to even number)", () => {
-		const messages: Anthropic.Messages.MessageParam[] = [
+		const messages: ApiMessage[] = [
 			{ role: "user", content: "First message" },
 			{ role: "assistant", content: "Second message" },
 			{ role: "user", content: "Third message" },
@@ -77,7 +78,7 @@ describe("truncateConversation", () => {
 	})
 
 	it("should round to an even number of messages to remove", () => {
-		const messages: Anthropic.Messages.MessageParam[] = [
+		const messages: ApiMessage[] = [
 			{ role: "user", content: "First message" },
 			{ role: "assistant", content: "Second message" },
 			{ role: "user", content: "Third message" },
@@ -96,7 +97,7 @@ describe("truncateConversation", () => {
 	})
 
 	it("should handle edge case with fracToRemove = 0", () => {
-		const messages: Anthropic.Messages.MessageParam[] = [
+		const messages: ApiMessage[] = [
 			{ role: "user", content: "First message" },
 			{ role: "assistant", content: "Second message" },
 			{ role: "user", content: "Third message" },
@@ -108,7 +109,7 @@ describe("truncateConversation", () => {
 	})
 
 	it("should handle edge case with fracToRemove = 1", () => {
-		const messages: Anthropic.Messages.MessageParam[] = [
+		const messages: ApiMessage[] = [
 			{ role: "user", content: "First message" },
 			{ role: "assistant", content: "Second message" },
 			{ role: "user", content: "Third message" },
@@ -224,7 +225,7 @@ describe("truncateConversationIfNeeded", () => {
 		maxTokens,
 	})
 
-	const messages: Anthropic.Messages.MessageParam[] = [
+	const messages: ApiMessage[] = [
 		{ role: "user", content: "First message" },
 		{ role: "assistant", content: "Second message" },
 		{ role: "user", content: "Third message" },
@@ -328,7 +329,7 @@ describe("truncateConversationIfNeeded", () => {
 		// Test case 1: Small content that won't push us over the threshold
 		const smallContent = [{ type: "text" as const, text: "Small content" }]
 		const smallContentTokens = await estimateTokenCount(smallContent, mockApiHandler)
-		const messagesWithSmallContent: Anthropic.Messages.MessageParam[] = [
+		const messagesWithSmallContent: ApiMessage[] = [
 			...messages.slice(0, -1),
 			{ role: messages[messages.length - 1].role, content: smallContent },
 		]
@@ -353,7 +354,7 @@ describe("truncateConversationIfNeeded", () => {
 			},
 		]
 		const largeContentTokens = await estimateTokenCount(largeContent, mockApiHandler)
-		const messagesWithLargeContent: Anthropic.Messages.MessageParam[] = [
+		const messagesWithLargeContent: ApiMessage[] = [
 			...messages.slice(0, -1),
 			{ role: messages[messages.length - 1].role, content: largeContent },
 		]
@@ -372,7 +373,7 @@ describe("truncateConversationIfNeeded", () => {
 		// Test case 3: Very large content that will definitely exceed threshold
 		const veryLargeContent = [{ type: "text" as const, text: "X".repeat(1000) }]
 		const veryLargeContentTokens = await estimateTokenCount(veryLargeContent, mockApiHandler)
-		const messagesWithVeryLargeContent: Anthropic.Messages.MessageParam[] = [
+		const messagesWithVeryLargeContent: ApiMessage[] = [
 			...messages.slice(0, -1),
 			{ role: messages[messages.length - 1].role, content: veryLargeContent },
 		]
@@ -424,7 +425,7 @@ describe("getMaxTokens", () => {
 	})
 
 	// Reuse across tests for consistency
-	const messages: Anthropic.Messages.MessageParam[] = [
+	const messages: ApiMessage[] = [
 		{ role: "user", content: "First message" },
 		{ role: "assistant", content: "Second message" },
 		{ role: "user", content: "Third message" },
diff --git a/src/core/sliding-window/index.ts b/src/core/sliding-window/index.ts
index 75395ecd758..d17bf7fc57d 100644
--- a/src/core/sliding-window/index.ts
+++ b/src/core/sliding-window/index.ts
@@ -1,5 +1,7 @@
 import { Anthropic } from "@anthropic-ai/sdk"
 import { ApiHandler } from "../../api"
+import { summarizeConversation } from "../condense"
+import { ApiMessage } from "../task-persistence/apiMessages"
 
 /**
  * Default percentage of the context window to use as a buffer when deciding when to truncate
@@ -27,14 +29,11 @@ export async function estimateTokenCount(
  * The first message is always retained, and a specified fraction (rounded to an even number)
  * of messages from the beginning (excluding the first) is removed.
  *
- * @param {Anthropic.Messages.MessageParam[]} messages - The conversation messages.
+ * @param {ApiMessage[]} messages - The conversation messages.
  * @param {number} fracToRemove - The fraction (between 0 and 1) of messages (excluding the first) to remove.
- * @returns {Anthropic.Messages.MessageParam[]} The truncated conversation messages.
+ * @returns {ApiMessage[]} The truncated conversation messages.
  */
-export function truncateConversation(
-	messages: Anthropic.Messages.MessageParam[],
-	fracToRemove: number,
-): Anthropic.Messages.MessageParam[] {
+export function truncateConversation(messages: ApiMessage[], fracToRemove: number): ApiMessage[] {
 	const truncatedMessages = [messages[0]]
 	const rawMessagesToRemove = Math.floor((messages.length - 1) * fracToRemove)
 	const messagesToRemove = rawMessagesToRemove - (rawMessagesToRemove % 2)
@@ -48,20 +47,22 @@ export function truncateConversation(
  * Conditionally truncates the conversation messages if the total token count
  * exceeds the model's limit, considering the size of incoming content.
  *
- * @param {Anthropic.Messages.MessageParam[]} messages - The conversation messages.
+ * @param {ApiMessage[]} messages - The conversation messages.
  * @param {number} totalTokens - The total number of tokens in the conversation (excluding the last user message).
  * @param {number} contextWindow - The context window size.
  * @param {number} maxTokens - The maximum number of tokens allowed.
  * @param {ApiHandler} apiHandler - The API handler to use for token counting.
- * @returns {Anthropic.Messages.MessageParam[]} The original or truncated conversation messages.
+ * @param {boolean} autoCondenseContext - Whether to use LLM summarization or sliding window implementation
+ * @returns {ApiMessage[]} The original or truncated conversation messages.
  */
 
 type TruncateOptions = {
-	messages: Anthropic.Messages.MessageParam[]
+	messages: ApiMessage[]
 	totalTokens: number
 	contextWindow: number
 	maxTokens?: number | null
 	apiHandler: ApiHandler
+	autoCondenseContext?: boolean
 }
 
 /**
@@ -69,7 +70,7 @@ type TruncateOptions = {
  * exceeds the model's limit, considering the size of incoming content.
  *
  * @param {TruncateOptions} options - The options for truncation
- * @returns {Promise<Anthropic.Messages.MessageParam[]>} The original or truncated conversation messages.
+ * @returns {Promise<ApiMessage[]>} The original or truncated conversation messages.
  */
 export async function truncateConversationIfNeeded({
 	messages,
@@ -77,7 +78,8 @@ export async function truncateConversationIfNeeded({
 	contextWindow,
 	maxTokens,
 	apiHandler,
-}: TruncateOptions): Promise<Anthropic.Messages.MessageParam[]> {
+	autoCondenseContext,
+}: TruncateOptions): Promise<ApiMessage[]> {
 	// Calculate the maximum tokens reserved for response
 	const reservedTokens = maxTokens || contextWindow * 0.2
 
@@ -96,5 +98,13 @@ export async function truncateConversationIfNeeded({
 	const allowedTokens = contextWindow * (1 - TOKEN_BUFFER_PERCENTAGE) - reservedTokens
 
 	// Determine if truncation is needed and apply if necessary
-	return effectiveTokens > allowedTokens ? truncateConversation(messages, 0.5) : messages
+	if (effectiveTokens <= allowedTokens) {
+		return messages
+	} else if (autoCondenseContext) {
+		const summarizedMessages = await summarizeConversation(messages, apiHandler)
+		if (messages !== summarizedMessages) {
+			return summarizedMessages
+		}
+	}
+	return truncateConversation(messages, 0.5)
 }
diff --git a/src/core/task-persistence/apiMessages.ts b/src/core/task-persistence/apiMessages.ts
index b361016345b..6ac36ed08fa 100644
--- a/src/core/task-persistence/apiMessages.ts
+++ b/src/core/task-persistence/apiMessages.ts
@@ -8,7 +8,7 @@ import { fileExistsAtPath } from "../../utils/fs"
 import { GlobalFileNames } from "../../shared/globalFileNames"
 import { getTaskDirectoryPath } from "../../shared/storagePathManager"
 
-export type ApiMessage = Anthropic.MessageParam & { ts?: number }
+export type ApiMessage = Anthropic.MessageParam & { ts?: number; isSummary?: boolean }
 
 export async function readApiMessages({
 	taskId,
diff --git a/src/core/task/Task.ts b/src/core/task/Task.ts
index 2d2180e391e..1b63ba9f4cc 100644
--- a/src/core/task/Task.ts
+++ b/src/core/task/Task.ts
@@ -79,6 +79,9 @@ import {
 	checkpointDiff,
 } from "../checkpoints"
 import { processUserContentMentions } from "../mentions/processUserContentMentions"
+import { ApiMessage } from "../task-persistence/apiMessages"
+import { getMessagesSinceLastSummary } from "../condense"
+import { maybeRemoveImageBlocks } from "../../api/transform/image-cleaning"
 
 export type ClineEvents = {
 	message: [{ action: "created" | "updated"; message: ClineMessage }]
@@ -155,7 +158,7 @@ export class Task extends EventEmitter<ClineEvents> {
 	didEditFile: boolean = false
 
 	// LLM Messages & Chat Messages
-	apiConversationHistory: (Anthropic.MessageParam & { ts?: number })[] = []
+	apiConversationHistory: ApiMessage[] = []
 	clineMessages: ClineMessage[] = []
 
 	// Ask
@@ -284,7 +287,7 @@ export class Task extends EventEmitter<ClineEvents> {
 
 	// API Messages
 
-	private async getSavedApiConversationHistory(): Promise<Anthropic.MessageParam[]> {
+	private async getSavedApiConversationHistory(): Promise<ApiMessage[]> {
 		return readApiMessages({ taskId: this.taskId, globalStoragePath: this.globalStoragePath })
 	}
 
@@ -294,7 +297,7 @@ export class Task extends EventEmitter<ClineEvents> {
 		await this.saveApiConversationHistory()
 	}
 
-	async overwriteApiConversationHistory(newHistory: Anthropic.MessageParam[]) {
+	async overwriteApiConversationHistory(newHistory: ApiMessage[]) {
 		this.apiConversationHistory = newHistory
 		await this.saveApiConversationHistory()
 	}
@@ -697,8 +700,7 @@ export class Task extends EventEmitter<ClineEvents> {
 
 		// Make sure that the api conversation history can be resumed by the API,
 		// even if it goes out of sync with cline messages.
-		let existingApiConversationHistory: Anthropic.Messages.MessageParam[] =
-			await this.getSavedApiConversationHistory()
+		let existingApiConversationHistory: ApiMessage[] = await this.getSavedApiConversationHistory()
 
 		// v2.0 xml tags refactor caveat: since we don't use tools anymore, we need to replace all tool use blocks with a text block since the API disallows conversations with tool uses and no tool schema
 		const conversationWithoutToolBlocks = existingApiConversationHistory.map((message) => {
@@ -742,7 +744,7 @@ export class Task extends EventEmitter<ClineEvents> {
 		// if the last message is a user message, we can need to get the assistant message before it to see if it made tool calls, and if so, fill in the remaining tool responses with 'interrupted'
 
 		let modifiedOldUserContent: Anthropic.Messages.ContentBlockParam[] // either the last message if its user message, or the user message before the last (assistant) message
-		let modifiedApiConversationHistory: Anthropic.Messages.MessageParam[] // need to remove the last user message to replace with new modified user message
+		let modifiedApiConversationHistory: ApiMessage[] // need to remove the last user message to replace with new modified user message
 		if (existingApiConversationHistory.length > 0) {
 			const lastMessage = existingApiConversationHistory[existingApiConversationHistory.length - 1]
 
@@ -768,7 +770,7 @@ export class Task extends EventEmitter<ClineEvents> {
 					modifiedOldUserContent = []
 				}
 			} else if (lastMessage.role === "user") {
-				const previousAssistantMessage: Anthropic.Messages.MessageParam | undefined =
+				const previousAssistantMessage: ApiMessage | undefined =
 					existingApiConversationHistory[existingApiConversationHistory.length - 2]
 
 				const existingUserContent: Anthropic.Messages.ContentBlockParam[] = Array.isArray(lastMessage.content)
@@ -1480,44 +1482,24 @@ export class Task extends EventEmitter<ClineEvents> {
 
 			const contextWindow = modelInfo.contextWindow
 
+			const autoCondenseContext = experiments?.autoCondenseContext ?? false
 			const trimmedMessages = await truncateConversationIfNeeded({
 				messages: this.apiConversationHistory,
 				totalTokens,
 				maxTokens,
 				contextWindow,
 				apiHandler: this.api,
+				autoCondenseContext,
 			})
-
 			if (trimmedMessages !== this.apiConversationHistory) {
 				await this.overwriteApiConversationHistory(trimmedMessages)
 			}
 		}
 
-		// Clean conversation history by:
-		// 1. Converting to Anthropic.MessageParam by spreading only the API-required properties.
-		// 2. Converting image blocks to text descriptions if model doesn't support images.
-		const cleanConversationHistory = this.apiConversationHistory.map(({ role, content }) => {
-			// Handle array content (could contain image blocks).
-			if (Array.isArray(content)) {
-				if (!this.api.getModel().info.supportsImages) {
-					// Convert image blocks to text descriptions.
-					content = content.map((block) => {
-						if (block.type === "image") {
-							// Convert image blocks to text descriptions.
-							// Note: We can't access the actual image content/url due to API limitations,
-							// but we can indicate that an image was present in the conversation.
-							return {
-								type: "text",
-								text: "[Referenced image in conversation]",
-							}
-						}
-						return block
-					})
-				}
-			}
-
-			return { role, content }
-		})
+		const messagesSinceLastSummary = getMessagesSinceLastSummary(this.apiConversationHistory)
+		const cleanConversationHistory = maybeRemoveImageBlocks(messagesSinceLastSummary, this.api).map(
+			({ role, content }) => ({ role, content }),
+		)
 
 		const stream = this.api.createMessage(systemPrompt, cleanConversationHistory)
 		const iterator = stream[Symbol.asyncIterator]()
diff --git a/src/exports/roo-code.d.ts b/src/exports/roo-code.d.ts
index 98290336ce5..d8ee8edd349 100644
--- a/src/exports/roo-code.d.ts
+++ b/src/exports/roo-code.d.ts
@@ -100,6 +100,7 @@ type GlobalSettings = {
 	fuzzyMatchThreshold?: number | undefined
 	experiments?:
 		| {
+				autoCondenseContext: boolean
 				powerSteering: boolean
 		  }
 		| undefined
@@ -786,6 +787,7 @@ type IpcMessage =
 								terminalCompressProgressBar?: boolean | undefined
 								experiments?:
 									| {
+											autoCondenseContext: boolean
 											powerSteering: boolean
 									  }
 									| undefined
@@ -1248,6 +1250,7 @@ type TaskCommand =
 					terminalCompressProgressBar?: boolean | undefined
 					experiments?:
 						| {
+								autoCondenseContext: boolean
 								powerSteering: boolean
 						  }
 						| undefined
diff --git a/src/exports/types.ts b/src/exports/types.ts
index 4b71ad5d0b0..87e9af06036 100644
--- a/src/exports/types.ts
+++ b/src/exports/types.ts
@@ -100,6 +100,7 @@ type GlobalSettings = {
 	fuzzyMatchThreshold?: number | undefined
 	experiments?:
 		| {
+				autoCondenseContext: boolean
 				powerSteering: boolean
 		  }
 		| undefined
@@ -798,6 +799,7 @@ type IpcMessage =
 								terminalCompressProgressBar?: boolean | undefined
 								experiments?:
 									| {
+											autoCondenseContext: boolean
 											powerSteering: boolean
 									  }
 									| undefined
@@ -1262,6 +1264,7 @@ type TaskCommand =
 					terminalCompressProgressBar?: boolean | undefined
 					experiments?:
 						| {
+								autoCondenseContext: boolean
 								powerSteering: boolean
 						  }
 						| undefined
diff --git a/src/schemas/index.ts b/src/schemas/index.ts
index 70ea39f6246..4ac5c8806bd 100644
--- a/src/schemas/index.ts
+++ b/src/schemas/index.ts
@@ -312,7 +312,7 @@ export type CommandExecutionStatus = z.infer<typeof commandExecutionStatusSchema
  * ExperimentId
  */
 
-export const experimentIds = ["powerSteering"] as const
+export const experimentIds = ["autoCondenseContext", "powerSteering"] as const
 
 export const experimentIdsSchema = z.enum(experimentIds)
 
@@ -323,6 +323,7 @@ export type ExperimentId = z.infer<typeof experimentIdsSchema>
  */
 
 const experimentsSchema = z.object({
+	autoCondenseContext: z.boolean(),
 	powerSteering: z.boolean(),
 })
 
diff --git a/src/shared/__tests__/experiments.test.ts b/src/shared/__tests__/experiments.test.ts
index b68de1ced12..9d0e6dab9c6 100644
--- a/src/shared/__tests__/experiments.test.ts
+++ b/src/shared/__tests__/experiments.test.ts
@@ -10,17 +10,28 @@ describe("experiments", () => {
 		})
 	})
 
+	describe("AUTO_CONDENSE_CONTEXT", () => {
+		it("is configured correctly", () => {
+			expect(EXPERIMENT_IDS.AUTO_CONDENSE_CONTEXT).toBe("autoCondenseContext")
+			expect(experimentConfigsMap.AUTO_CONDENSE_CONTEXT).toMatchObject({
+				enabled: false,
+			})
+		})
+	})
+
 	describe("isEnabled", () => {
-		it("returns false when experiment is not enabled", () => {
+		it("returns false when POWER_STEERING experiment is not enabled", () => {
 			const experiments: Record<ExperimentId, boolean> = {
 				powerSteering: false,
+				autoCondenseContext: false,
 			}
 			expect(Experiments.isEnabled(experiments, EXPERIMENT_IDS.POWER_STEERING)).toBe(false)
 		})
 
-		it("returns true when experiment is enabled", () => {
+		it("returns true when experiment POWER_STEERING is enabled", () => {
 			const experiments: Record<ExperimentId, boolean> = {
 				powerSteering: true,
+				autoCondenseContext: false,
 			}
 			expect(Experiments.isEnabled(experiments, EXPERIMENT_IDS.POWER_STEERING)).toBe(true)
 		})
@@ -28,8 +39,25 @@ describe("experiments", () => {
 		it("returns false when experiment is not present", () => {
 			const experiments: Record<ExperimentId, boolean> = {
 				powerSteering: false,
+				autoCondenseContext: false,
 			}
 			expect(Experiments.isEnabled(experiments, EXPERIMENT_IDS.POWER_STEERING)).toBe(false)
 		})
+
+		it("returns false when AUTO_CONDENSE_CONTEXT experiment is not enabled", () => {
+			const experiments: Record<ExperimentId, boolean> = {
+				powerSteering: false,
+				autoCondenseContext: false,
+			}
+			expect(Experiments.isEnabled(experiments, EXPERIMENT_IDS.AUTO_CONDENSE_CONTEXT)).toBe(false)
+		})
+
+		it("returns true when AUTO_CONDENSE_CONTEXT experiment is enabled", () => {
+			const experiments: Record<ExperimentId, boolean> = {
+				powerSteering: false,
+				autoCondenseContext: true,
+			}
+			expect(Experiments.isEnabled(experiments, EXPERIMENT_IDS.AUTO_CONDENSE_CONTEXT)).toBe(true)
+		})
 	})
 })
diff --git a/src/shared/experiments.ts b/src/shared/experiments.ts
index b8917d8b108..0c9cba9c670 100644
--- a/src/shared/experiments.ts
+++ b/src/shared/experiments.ts
@@ -4,6 +4,7 @@ import { AssertEqual, Equals, Keys, Values } from "../utils/type-fu"
 export type { ExperimentId }
 
 export const EXPERIMENT_IDS = {
+	AUTO_CONDENSE_CONTEXT: "autoCondenseContext",
 	POWER_STEERING: "powerSteering",
 } as const satisfies Record<string, ExperimentId>
 
@@ -16,6 +17,7 @@ interface ExperimentConfig {
 }
 
 export const experimentConfigsMap: Record<ExperimentKey, ExperimentConfig> = {
+	AUTO_CONDENSE_CONTEXT: { enabled: false },
 	POWER_STEERING: { enabled: false },
 }
 
diff --git a/webview-ui/src/context/__tests__/ExtensionStateContext.test.tsx b/webview-ui/src/context/__tests__/ExtensionStateContext.test.tsx
index 02e57b5e354..dbac052944f 100644
--- a/webview-ui/src/context/__tests__/ExtensionStateContext.test.tsx
+++ b/webview-ui/src/context/__tests__/ExtensionStateContext.test.tsx
@@ -215,6 +215,7 @@ describe("mergeExtensionState", () => {
 			apiConfiguration: { modelMaxThinkingTokens: 456, modelTemperature: 0.3 },
 			experiments: {
 				powerSteering: true,
+				autoCondenseContext: true,
 			} as Record<ExperimentId, boolean>,
 		}
 
@@ -227,6 +228,7 @@ describe("mergeExtensionState", () => {
 
 		expect(result.experiments).toEqual({
 			powerSteering: true,
+			autoCondenseContext: true,
 		})
 	})
 })
diff --git a/webview-ui/src/i18n/locales/ca/settings.json b/webview-ui/src/i18n/locales/ca/settings.json
index c9e1cb5606a..a6262f5c0d3 100644
--- a/webview-ui/src/i18n/locales/ca/settings.json
+++ b/webview-ui/src/i18n/locales/ca/settings.json
@@ -396,6 +396,10 @@
 	},
 	"experimental": {
 		"warning": "⚠️",
+		"AUTO_CONDENSE_CONTEXT": {
+			"name": "Condensar intel·ligentment la finestra de context",
+			"description": "Utilitza una crida LLM per resumir la conversa anterior quan la finestra de context de la tasca està gairebé plena, en lloc d'eliminar missatges antics. Avís: el cost de resumir actualment no s'inclou en els costos d'API mostrats a la interfície."
+		},
 		"DIFF_STRATEGY_UNIFIED": {
 			"name": "Utilitzar estratègia diff unificada experimental",
 			"description": "Activar l'estratègia diff unificada experimental. Aquesta estratègia podria reduir el nombre de reintents causats per errors del model, però pot causar comportaments inesperats o edicions incorrectes. Activeu-la només si enteneu els riscos i esteu disposats a revisar acuradament tots els canvis."
diff --git a/webview-ui/src/i18n/locales/de/settings.json b/webview-ui/src/i18n/locales/de/settings.json
index aac308c4a1a..c15e0955b3e 100644
--- a/webview-ui/src/i18n/locales/de/settings.json
+++ b/webview-ui/src/i18n/locales/de/settings.json
@@ -396,6 +396,10 @@
 	},
 	"experimental": {
 		"warning": "⚠️",
+		"AUTO_CONDENSE_CONTEXT": {
+			"name": "Kontextfenster intelligent komprimieren",
+			"description": "Verwendet einen LLM-Aufruf, um das vorherige Gespräch zusammenzufassen, wenn das Kontextfenster der Aufgabe fast voll ist, anstatt alte Nachrichten zu verwerfen. Hinweis: Die Kosten für die Zusammenfassung sind derzeit nicht in den in der Benutzeroberfläche angezeigten API-Kosten enthalten."
+		},
 		"DIFF_STRATEGY_UNIFIED": {
 			"name": "Experimentelle einheitliche Diff-Strategie verwenden",
 			"description": "Aktiviert die experimentelle einheitliche Diff-Strategie. Diese Strategie könnte die Anzahl der durch Modellfehler verursachten Wiederholungen reduzieren, kann aber unerwartetes Verhalten oder falsche Bearbeitungen verursachen. Nur aktivieren, wenn du die Risiken verstehst und bereit bist, alle Änderungen sorgfältig zu überprüfen."
diff --git a/webview-ui/src/i18n/locales/en/settings.json b/webview-ui/src/i18n/locales/en/settings.json
index 12d94b43375..84d6c2ef368 100644
--- a/webview-ui/src/i18n/locales/en/settings.json
+++ b/webview-ui/src/i18n/locales/en/settings.json
@@ -396,6 +396,10 @@
 	},
 	"experimental": {
 		"warning": "⚠️",
+		"AUTO_CONDENSE_CONTEXT": {
+			"name": "Intelligently condense the context window",
+			"description": "Uses an LLM call to summarize the past conversation when the task's context window is almost full, rather than dropping old messages. Disclaimer: the cost of summarizing is not currently included in the API costs shown in the UI."
+		},
 		"DIFF_STRATEGY_UNIFIED": {
 			"name": "Use experimental unified diff strategy",
 			"description": "Enable the experimental unified diff strategy. This strategy might reduce the number of retries caused by model errors but may cause unexpected behavior or incorrect edits. Only enable if you understand the risks and are willing to carefully review all changes."
diff --git a/webview-ui/src/i18n/locales/es/settings.json b/webview-ui/src/i18n/locales/es/settings.json
index f862f134055..26047fadbfb 100644
--- a/webview-ui/src/i18n/locales/es/settings.json
+++ b/webview-ui/src/i18n/locales/es/settings.json
@@ -396,6 +396,10 @@
 	},
 	"experimental": {
 		"warning": "⚠️",
+		"AUTO_CONDENSE_CONTEXT": {
+			"name": "Condensar inteligentemente la ventana de contexto",
+			"description": "Utiliza una llamada LLM para resumir la conversación anterior cuando la ventana de contexto de la tarea está casi llena, en lugar de eliminar mensajes antiguos. Aviso: el costo de resumir actualmente no está incluido en los costos de API mostrados en la interfaz."
+		},
 		"DIFF_STRATEGY_UNIFIED": {
 			"name": "Usar estrategia de diff unificada experimental",
 			"description": "Habilitar la estrategia de diff unificada experimental. Esta estrategia podría reducir el número de reintentos causados por errores del modelo, pero puede causar comportamientos inesperados o ediciones incorrectas. Habilítela solo si comprende los riesgos y está dispuesto a revisar cuidadosamente todos los cambios."
diff --git a/webview-ui/src/i18n/locales/fr/settings.json b/webview-ui/src/i18n/locales/fr/settings.json
index 9f694c5aa33..a5b6a48d78a 100644
--- a/webview-ui/src/i18n/locales/fr/settings.json
+++ b/webview-ui/src/i18n/locales/fr/settings.json
@@ -396,6 +396,10 @@
 	},
 	"experimental": {
 		"warning": "⚠️",
+		"AUTO_CONDENSE_CONTEXT": {
+			"name": "Condenser intelligemment la fenêtre de contexte",
+			"description": "Utilise un appel LLM pour résumer la conversation précédente lorsque la fenêtre de contexte de la tâche est presque pleine, plutôt que de supprimer les anciens messages. Avertissement : le coût de la synthèse n'est actuellement pas inclus dans les coûts API affichés dans l'interface."
+		},
 		"DIFF_STRATEGY_UNIFIED": {
 			"name": "Utiliser la stratégie diff unifiée expérimentale",
 			"description": "Activer la stratégie diff unifiée expérimentale. Cette stratégie pourrait réduire le nombre de tentatives causées par des erreurs de modèle, mais peut provoquer des comportements inattendus ou des modifications incorrectes. Activez-la uniquement si vous comprenez les risques et êtes prêt à examiner attentivement tous les changements."
diff --git a/webview-ui/src/i18n/locales/hi/settings.json b/webview-ui/src/i18n/locales/hi/settings.json
index 0b21ed906b2..aa5efaaf87a 100644
--- a/webview-ui/src/i18n/locales/hi/settings.json
+++ b/webview-ui/src/i18n/locales/hi/settings.json
@@ -396,6 +396,10 @@
 	},
 	"experimental": {
 		"warning": "⚠️",
+		"AUTO_CONDENSE_CONTEXT": {
+			"name": "संदर्भ विंडो को बुद्धिमानी से संघनित करें",
+			"description": "जब कार्य का संदर्भ विंडो लगभग भर जाता है, तो पुराने संदेशों को हटाने के बजाय पिछली बातचीत को संक्षेप में प्रस्तुत करने के लिए LLM कॉल का उपयोग करता है। अस्वीकरण: संक्षेपण की लागत वर्तमान में UI में दिखाए गए API लागतों में शामिल नहीं है।"
+		},
 		"DIFF_STRATEGY_UNIFIED": {
 			"name": "प्रायोगिक एकीकृत diff रणनीति का उपयोग करें",
 			"description": "प्रायोगिक एकीकृत diff रणनीति सक्षम करें। यह रणनीति मॉडल त्रुटियों के कारण पुनः प्रयासों की संख्या को कम कर सकती है, लेकिन अप्रत्याशित व्यवहार या गलत संपादन का कारण बन सकती है। केवल तभी सक्षम करें जब आप जोखिमों को समझते हों और सभी परिवर्तनों की सावधानीपूर्वक समीक्षा करने के लिए तैयार हों।"
diff --git a/webview-ui/src/i18n/locales/it/settings.json b/webview-ui/src/i18n/locales/it/settings.json
index a3e210c4808..67dc5c77b1e 100644
--- a/webview-ui/src/i18n/locales/it/settings.json
+++ b/webview-ui/src/i18n/locales/it/settings.json
@@ -396,6 +396,10 @@
 	},
 	"experimental": {
 		"warning": "⚠️",
+		"AUTO_CONDENSE_CONTEXT": {
+			"name": "Condensa intelligentemente la finestra di contesto",
+			"description": "Utilizza una chiamata LLM per riassumere la conversazione precedente quando la finestra di contesto dell'attività è quasi piena, invece di eliminare i messaggi vecchi. Avviso: il costo della sintesi non è attualmente incluso nei costi API mostrati nell'interfaccia."
+		},
 		"DIFF_STRATEGY_UNIFIED": {
 			"name": "Usa strategia diff unificata sperimentale",
 			"description": "Abilita la strategia diff unificata sperimentale. Questa strategia potrebbe ridurre il numero di tentativi causati da errori del modello, ma può causare comportamenti imprevisti o modifiche errate. Abilitala solo se comprendi i rischi e sei disposto a rivedere attentamente tutte le modifiche."
diff --git a/webview-ui/src/i18n/locales/ja/settings.json b/webview-ui/src/i18n/locales/ja/settings.json
index 1f1ec12c54a..4cf8f6d16e1 100644
--- a/webview-ui/src/i18n/locales/ja/settings.json
+++ b/webview-ui/src/i18n/locales/ja/settings.json
@@ -396,6 +396,10 @@
 	},
 	"experimental": {
 		"warning": "⚠️",
+		"AUTO_CONDENSE_CONTEXT": {
+			"name": "コンテキストウィンドウをインテリジェントに圧縮する",
+			"description": "タスクのコンテキストウィンドウがほぼいっぱいになったとき、古いメッセージを削除する代わりに、LLM呼び出しを使用して過去の会話を要約します。免責事項：要約のコストは現在UIに表示されるAPIコストには含まれていません。"
+		},
 		"DIFF_STRATEGY_UNIFIED": {
 			"name": "実験的な統合diff戦略を使用する",
 			"description": "実験的な統合diff戦略を有効にします。この戦略はモデルエラーによる再試行の回数を減らす可能性がありますが、予期しない動作や不正確な編集を引き起こす可能性があります。リスクを理解し、すべての変更を注意深く確認する準備がある場合にのみ有効にしてください。"
diff --git a/webview-ui/src/i18n/locales/ko/settings.json b/webview-ui/src/i18n/locales/ko/settings.json
index 95195ff0a23..72c0e160a18 100644
--- a/webview-ui/src/i18n/locales/ko/settings.json
+++ b/webview-ui/src/i18n/locales/ko/settings.json
@@ -396,6 +396,10 @@
 	},
 	"experimental": {
 		"warning": "⚠️",
+		"AUTO_CONDENSE_CONTEXT": {
+			"name": "컨텍스트 창을 지능적으로 압축",
+			"description": "작업의 컨텍스트 창이 거의 가득 찼을 때 이전 메시지를 삭제하는 대신 LLM 호출을 사용하여 이전 대화를 요약합니다. 참고: 요약 비용은 현재 UI에 표시된 API 비용에 포함되지 않습니다."
+		},
 		"DIFF_STRATEGY_UNIFIED": {
 			"name": "실험적 통합 diff 전략 사용",
 			"description": "실험적 통합 diff 전략을 활성화합니다. 이 전략은 모델 오류로 인한 재시도 횟수를 줄일 수 있지만 예기치 않은 동작이나 잘못된 편집을 일으킬 수 있습니다. 위험을 이해하고 모든 변경 사항을 신중하게 검토할 의향이 있는 경우에만 활성화하십시오."
diff --git a/webview-ui/src/i18n/locales/nl/settings.json b/webview-ui/src/i18n/locales/nl/settings.json
index e812d78f26b..54ecc15637c 100644
--- a/webview-ui/src/i18n/locales/nl/settings.json
+++ b/webview-ui/src/i18n/locales/nl/settings.json
@@ -396,6 +396,10 @@
 	},
 	"experimental": {
 		"warning": "⚠️",
+		"AUTO_CONDENSE_CONTEXT": {
+			"name": "Contextvenster intelligent comprimeren",
+			"description": "Gebruikt een LLM-aanroep om eerdere gesprekken samen te vatten wanneer het contextvenster van de taak bijna vol is, in plaats van oude berichten te verwijderen. Let op: de kosten van het samenvatten zijn momenteel niet inbegrepen in de API-kosten die in de interface worden getoond."
+		},
 		"DIFF_STRATEGY_UNIFIED": {
 			"name": "Experimentele unified diff-strategie gebruiken",
 			"description": "Schakel de experimentele unified diff-strategie in. Deze strategie kan het aantal herhalingen door model fouten verminderen, maar kan onverwacht gedrag of onjuiste bewerkingen veroorzaken. Alleen inschakelen als je de risico's begrijpt en wijzigingen zorgvuldig wilt controleren."
diff --git a/webview-ui/src/i18n/locales/pl/settings.json b/webview-ui/src/i18n/locales/pl/settings.json
index 343ce013977..14b3a1161b8 100644
--- a/webview-ui/src/i18n/locales/pl/settings.json
+++ b/webview-ui/src/i18n/locales/pl/settings.json
@@ -396,6 +396,10 @@
 	},
 	"experimental": {
 		"warning": "⚠️",
+		"AUTO_CONDENSE_CONTEXT": {
+			"name": "Inteligentnie kondensuj okno kontekstu",
+			"description": "Używa wywołania LLM do podsumowania wcześniejszej rozmowy, gdy okno kontekstu zadania jest prawie pełne, zamiast usuwać stare wiadomości. Zastrzeżenie: koszt podsumowania nie jest obecnie uwzględniony w kosztach API pokazywanych w interfejsie."
+		},
 		"DIFF_STRATEGY_UNIFIED": {
 			"name": "Użyj eksperymentalnej ujednoliconej strategii diff",
 			"description": "Włącz eksperymentalną ujednoliconą strategię diff. Ta strategia może zmniejszyć liczbę ponownych prób spowodowanych błędami modelu, ale może powodować nieoczekiwane zachowanie lub nieprawidłowe edycje. Włącz tylko jeśli rozumiesz ryzyko i jesteś gotów dokładnie przeglądać wszystkie zmiany."
diff --git a/webview-ui/src/i18n/locales/pt-BR/settings.json b/webview-ui/src/i18n/locales/pt-BR/settings.json
index 0c66a218477..eeaefd06e8b 100644
--- a/webview-ui/src/i18n/locales/pt-BR/settings.json
+++ b/webview-ui/src/i18n/locales/pt-BR/settings.json
@@ -396,6 +396,10 @@
 	},
 	"experimental": {
 		"warning": "⚠️",
+		"AUTO_CONDENSE_CONTEXT": {
+			"name": "Condensar inteligentemente a janela de contexto",
+			"description": "Usa uma chamada LLM para resumir a conversa anterior quando a janela de contexto da tarefa está quase cheia, em vez de descartar mensagens antigas. Aviso: o custo de resumir não está atualmente incluído nos custos de API mostrados na interface."
+		},
 		"DIFF_STRATEGY_UNIFIED": {
 			"name": "Usar estratégia diff unificada experimental",
 			"description": "Ativar a estratégia diff unificada experimental. Esta estratégia pode reduzir o número de novas tentativas causadas por erros do modelo, mas pode causar comportamento inesperado ou edições incorretas. Ative apenas se compreender os riscos e estiver disposto a revisar cuidadosamente todas as alterações."
diff --git a/webview-ui/src/i18n/locales/ru/settings.json b/webview-ui/src/i18n/locales/ru/settings.json
index 376763a7d81..66e74b13329 100644
--- a/webview-ui/src/i18n/locales/ru/settings.json
+++ b/webview-ui/src/i18n/locales/ru/settings.json
@@ -396,6 +396,10 @@
 	},
 	"experimental": {
 		"warning": "⚠️",
+		"AUTO_CONDENSE_CONTEXT": {
+			"name": "Интеллектуальное сжатие контекстного окна",
+			"description": "Использует вызов LLM для обобщения предыдущего разговора, когда контекстное окно задачи почти заполнено, вместо удаления старых сообщений. Примечание: стоимость обобщения в настоящее время не включена в стоимость API, отображаемую в интерфейсе."
+		},
 		"DIFF_STRATEGY_UNIFIED": {
 			"name": "Использовать экспериментальную стратегию унифицированного диффа",
 			"description": "Включает экспериментальную стратегию унифицированного диффа. Может уменьшить количество повторных попыток из-за ошибок модели, но может привести к неожиданному поведению или неверным правкам. Включайте только если готовы внимательно проверять все изменения."
diff --git a/webview-ui/src/i18n/locales/tr/settings.json b/webview-ui/src/i18n/locales/tr/settings.json
index 1d23b7e04e5..2856d2eeb21 100644
--- a/webview-ui/src/i18n/locales/tr/settings.json
+++ b/webview-ui/src/i18n/locales/tr/settings.json
@@ -396,6 +396,10 @@
 	},
 	"experimental": {
 		"warning": "⚠️",
+		"AUTO_CONDENSE_CONTEXT": {
+			"name": "Bağlam penceresini akıllıca sıkıştır",
+			"description": "Görevin bağlam penceresi neredeyse dolduğunda, eski mesajları atmak yerine önceki konuşmayı özetlemek için bir LLM çağrısı kullanır. Not: Özetleme maliyeti şu anda arayüzde gösterilen API maliyetlerine dahil değildir."
+		},
 		"DIFF_STRATEGY_UNIFIED": {
 			"name": "Deneysel birleştirilmiş diff stratejisini kullan",
 			"description": "Deneysel birleştirilmiş diff stratejisini etkinleştir. Bu strateji, model hatalarından kaynaklanan yeniden deneme sayısını azaltabilir, ancak beklenmeyen davranışlara veya hatalı düzenlemelere neden olabilir. Yalnızca riskleri anlıyorsanız ve tüm değişiklikleri dikkatlice incelemeye istekliyseniz etkinleştirin."
diff --git a/webview-ui/src/i18n/locales/vi/settings.json b/webview-ui/src/i18n/locales/vi/settings.json
index f964bf4b7ac..b2ce14552b2 100644
--- a/webview-ui/src/i18n/locales/vi/settings.json
+++ b/webview-ui/src/i18n/locales/vi/settings.json
@@ -396,6 +396,10 @@
 	},
 	"experimental": {
 		"warning": "⚠️",
+		"AUTO_CONDENSE_CONTEXT": {
+			"name": "Nén cửa sổ ngữ cảnh một cách thông minh",
+			"description": "Sử dụng một lệnh gọi LLM để tóm tắt cuộc trò chuyện trước đó khi cửa sổ ngữ cảnh của tác vụ gần đầy, thay vì loại bỏ các tin nhắn cũ. Lưu ý: chi phí tóm tắt hiện không được tính vào chi phí API hiển thị trong giao diện người dùng."
+		},
 		"DIFF_STRATEGY_UNIFIED": {
 			"name": "Sử dụng chiến lược diff thống nhất thử nghiệm",
 			"description": "Bật chiến lược diff thống nhất thử nghiệm. Chiến lược này có thể giảm số lần thử lại do lỗi mô hình nhưng có thể gây ra hành vi không mong muốn hoặc chỉnh sửa không chính xác. Chỉ bật nếu bạn hiểu rõ các rủi ro và sẵn sàng xem xét cẩn thận tất cả các thay đổi."
diff --git a/webview-ui/src/i18n/locales/zh-CN/settings.json b/webview-ui/src/i18n/locales/zh-CN/settings.json
index ba14e487621..cd42bc4bcf3 100644
--- a/webview-ui/src/i18n/locales/zh-CN/settings.json
+++ b/webview-ui/src/i18n/locales/zh-CN/settings.json
@@ -396,6 +396,10 @@
 	},
 	"experimental": {
 		"warning": "⚠️",
+		"AUTO_CONDENSE_CONTEXT": {
+			"name": "智能压缩上下文窗口",
+			"description": "当任务上下文窗口接近填满时，使用 LLM 调用来总结过去的对话，而不是删除旧消息。注意：目前 UI 中显示的 API 费用不包括总结的成本。"
+		},
 		"DIFF_STRATEGY_UNIFIED": {
 			"name": "启用diff更新工具",
 			"description": "可减少因模型错误导致的重复尝试，但可能引发意外操作。启用前请确保理解风险并会仔细检查所有修改。"
diff --git a/webview-ui/src/i18n/locales/zh-TW/settings.json b/webview-ui/src/i18n/locales/zh-TW/settings.json
index 78c994457eb..4d0ec4420dc 100644
--- a/webview-ui/src/i18n/locales/zh-TW/settings.json
+++ b/webview-ui/src/i18n/locales/zh-TW/settings.json
@@ -396,6 +396,10 @@
 	},
 	"experimental": {
 		"warning": "⚠️",
+		"AUTO_CONDENSE_CONTEXT": {
+			"name": "智慧壓縮上下文視窗",
+			"description": "當工作的上下文視窗接近填滿時，使用 LLM 呼叫來摘要過去的對話，而非捨棄舊訊息。注意：目前 UI 中顯示的 API 費用並未包含摘要的成本。"
+		},
 		"DIFF_STRATEGY_UNIFIED": {
 			"name": "使用實驗性統一差異比對策略",
 			"description": "啟用實驗性的統一差異比對策略。此策略可能減少因模型錯誤而導致的重試次數，但也可能導致意外行為或錯誤的編輯。請務必了解風險，並願意仔細檢查所有變更後再啟用。"