Enable parallel tool calls by default

packages/types/src/experiment.ts

@@ -6,13 +6,7 @@ import type { Keys, Equals, AssertEqual } from "./type-fu.js"
  * ExperimentId
  */
 
-export const experimentIds = [
-	"preventFocusDisruption",
-	"imageGeneration",
-	"runSlashCommand",
-	"multipleNativeToolCalls",
-	"customTools",
-] as const
+export const experimentIds = ["preventFocusDisruption", "imageGeneration", "runSlashCommand", "customTools"] as const
 
 export const experimentIdsSchema = z.enum(experimentIds)
 
@@ -26,7 +20,6 @@ export const experimentsSchema = z.object({
 	preventFocusDisruption: z.boolean().optional(),
 	imageGeneration: z.boolean().optional(),
 	runSlashCommand: z.boolean().optional(),
-	multipleNativeToolCalls: z.boolean().optional(),
 	customTools: z.boolean().optional(),
 })
 

src/api/index.ts

@@ -84,8 +84,8 @@ export interface ApiHandlerCreateMessageMetadata {
 	tool_choice?: OpenAI.Chat.ChatCompletionCreateParams["tool_choice"]
 	/**
 	 * Controls whether the model can return multiple tool calls in a single response.
-	 * When true, parallel tool calls are enabled (OpenAI's parallel_tool_calls=true).
-	 * When false (default), only one tool call is returned per response.
+	 * When true (default), parallel tool calls are enabled (OpenAI's parallel_tool_calls=true).
+	 * When false, only one tool call is returned per response.
 	 */
 	parallelToolCalls?: boolean
 	/**

src/api/providers/__tests__/anthropic-vertex.spec.ts

@@ -1197,7 +1197,7 @@ describe("VertexHandler", () => {
 							}),
 						}),
 					]),
-					tool_choice: { type: "auto", disable_parallel_tool_use: true },
+					tool_choice: { type: "auto", disable_parallel_tool_use: false },
 				}),
 				undefined,
 			)

src/api/providers/__tests__/anthropic.spec.ts

@@ -514,7 +514,7 @@ describe("AnthropicHandler", () => {
 
 			expect(mockCreate).toHaveBeenCalledWith(
 				expect.objectContaining({
-					tool_choice: { type: "auto", disable_parallel_tool_use: true },
+					tool_choice: { type: "auto", disable_parallel_tool_use: false },
 				}),
 				expect.anything(),
 			)
@@ -535,7 +535,7 @@ describe("AnthropicHandler", () => {
 
 			expect(mockCreate).toHaveBeenCalledWith(
 				expect.objectContaining({
-					tool_choice: { type: "any", disable_parallel_tool_use: true },
+					tool_choice: { type: "any", disable_parallel_tool_use: false },
 				}),
 				expect.anything(),
 			)
@@ -581,7 +581,7 @@ describe("AnthropicHandler", () => {
 
 			expect(mockCreate).toHaveBeenCalledWith(
 				expect.objectContaining({
-					tool_choice: { type: "tool", name: "get_weather", disable_parallel_tool_use: true },
+					tool_choice: { type: "tool", name: "get_weather", disable_parallel_tool_use: false },
 				}),
 				expect.anything(),
 			)

src/api/providers/__tests__/deepinfra.spec.ts

@@ -214,9 +214,9 @@ describe("DeepInfraHandler", () => {
 					]),
 				}),
 			)
-			// parallel_tool_calls should be false when not explicitly set
+			// parallel_tool_calls should be true by default when not explicitly set
 			const callArgs = mockCreate.mock.calls[0][0]
-			expect(callArgs).toHaveProperty("parallel_tool_calls", false)
+			expect(callArgs).toHaveProperty("parallel_tool_calls", true)
 		})
 
 		it("should include tool_choice when provided", async () => {
@@ -264,8 +264,8 @@ describe("DeepInfraHandler", () => {
 			// Tools are now always present (minimum 6 from ALWAYS_AVAILABLE_TOOLS)
 			expect(callArgs).toHaveProperty("tools")
 			expect(callArgs).toHaveProperty("tool_choice")
-			// parallel_tool_calls should be false when not explicitly set
-			expect(callArgs).toHaveProperty("parallel_tool_calls", false)
+			// parallel_tool_calls should be true by default when not explicitly set
+			expect(callArgs).toHaveProperty("parallel_tool_calls", true)
 		})
 
 		it("should yield tool_call_partial chunks during streaming", async () => {

src/api/providers/__tests__/lmstudio-native-tools.spec.ts

@@ -82,9 +82,9 @@ describe("LmStudioHandler Native Tools", () => {
 					]),
 				}),
 			)
-			// parallel_tool_calls should be false when not explicitly set
+			// parallel_tool_calls should be true by default when not explicitly set
 			const callArgs = mockCreate.mock.calls[0][0]
-			expect(callArgs).toHaveProperty("parallel_tool_calls", false)
+			expect(callArgs).toHaveProperty("parallel_tool_calls", true)
 		})
 
 		it("should include tool_choice when provided", async () => {
@@ -128,8 +128,8 @@ describe("LmStudioHandler Native Tools", () => {
 			// Tools are now always present (minimum 6 from ALWAYS_AVAILABLE_TOOLS)
 			expect(callArgs).toHaveProperty("tools")
 			expect(callArgs).toHaveProperty("tool_choice")
-			// parallel_tool_calls should be false when not explicitly set
-			expect(callArgs).toHaveProperty("parallel_tool_calls", false)
+			// parallel_tool_calls should be true by default when not explicitly set
+			expect(callArgs).toHaveProperty("parallel_tool_calls", true)
 		})
 
 		it("should yield tool_call_partial chunks during streaming", async () => {

src/api/providers/__tests__/openai-native-tools.spec.ts

@@ -61,7 +61,7 @@ describe("OpenAiHandler native tools", () => {
 						function: expect.objectContaining({ name: "test_tool" }),
 					}),
 				]),
-				parallel_tool_calls: false,
+				parallel_tool_calls: true,
 			}),
 			expect.anything(),
 		)

src/api/providers/__tests__/openai.spec.ts

@@ -635,7 +635,7 @@ describe("OpenAiHandler", () => {
 					temperature: 0,
 					tools: undefined,
 					tool_choice: undefined,
-					parallel_tool_calls: false,
+					parallel_tool_calls: true,
 				},
 				{ path: "/models/chat/completions" },
 			)
@@ -684,7 +684,7 @@ describe("OpenAiHandler", () => {
 					],
 					tools: undefined,
 					tool_choice: undefined,
-					parallel_tool_calls: false,
+					parallel_tool_calls: true,
 				},
 				{ path: "/models/chat/completions" },
 			)

src/api/providers/__tests__/qwen-code-native-tools.spec.ts

@@ -99,7 +99,7 @@ describe("QwenCodeHandler Native Tools", () => {
 							}),
 						}),
 					]),
-					parallel_tool_calls: false,
+					parallel_tool_calls: true,
 				}),
 			)
 		})
@@ -145,7 +145,7 @@ describe("QwenCodeHandler Native Tools", () => {
 			const callArgs = mockCreate.mock.calls[mockCreate.mock.calls.length - 1][0]
 			expect(callArgs).toHaveProperty("tools")
 			expect(callArgs).toHaveProperty("tool_choice")
-			expect(callArgs).toHaveProperty("parallel_tool_calls", false)
+			expect(callArgs).toHaveProperty("parallel_tool_calls", true)
 		})
 
 		it("should yield tool_call_partial chunks during streaming", async () => {

src/api/providers/__tests__/unbound.spec.ts

@@ -375,7 +375,7 @@ describe("UnboundHandler", () => {
 							}),
 						}),
 					]),
-					parallel_tool_calls: false,
+					parallel_tool_calls: true,
 				}),
 				expect.objectContaining({
 					headers: {
@@ -435,7 +435,7 @@ describe("UnboundHandler", () => {
 			const callArgs = mockCreate.mock.calls[mockCreate.mock.calls.length - 1][0]
 			expect(callArgs).toHaveProperty("tools")
 			expect(callArgs).toHaveProperty("tool_choice")
-			expect(callArgs).toHaveProperty("parallel_tool_calls", false)
+			expect(callArgs).toHaveProperty("parallel_tool_calls", true)
 		})
 
 		it("should yield tool_call_partial chunks during streaming", async () => {

src/api/providers/__tests__/vercel-ai-gateway.spec.ts

@@ -366,7 +366,7 @@ describe("VercelAiGatewayHandler", () => {
 				)
 			})
 
-			it("should include parallel_tool_calls: false by default", async () => {
+			it("should include parallel_tool_calls: true by default", async () => {
 				const handler = new VercelAiGatewayHandler(mockOptions)
 
 				const messageGenerator = handler.createMessage("test prompt", [], {
@@ -378,7 +378,7 @@ describe("VercelAiGatewayHandler", () => {
 				expect(mockCreate).toHaveBeenCalledWith(
 					expect.objectContaining({
 						tools: expect.any(Array),
-						parallel_tool_calls: false,
+						parallel_tool_calls: true,
 					}),
 				)
 			})

src/api/providers/__tests__/xai.spec.ts

@@ -339,7 +339,7 @@ describe("XAIHandler", () => {
 							}),
 						}),
 					]),
-					parallel_tool_calls: false,
+					parallel_tool_calls: true,
 				}),
 			)
 		})
@@ -393,7 +393,7 @@ describe("XAIHandler", () => {
 			const callArgs = mockCreate.mock.calls[mockCreate.mock.calls.length - 1][0]
 			expect(callArgs).toHaveProperty("tools")
 			expect(callArgs).toHaveProperty("tool_choice")
-			expect(callArgs).toHaveProperty("parallel_tool_calls", false)
+			expect(callArgs).toHaveProperty("parallel_tool_calls", true)
 		})
 
 		it("should yield tool_call_partial chunks during streaming", async () => {

src/api/providers/base-openai-compatible-provider.ts

@@ -95,7 +95,7 @@ export abstract class BaseOpenAiCompatibleProvider<ModelName extends string>
 			stream_options: { include_usage: true },
 			tools: this.convertToolsForOpenAI(metadata?.tools),
 			tool_choice: metadata?.tool_choice,
-			parallel_tool_calls: metadata?.parallelToolCalls ?? false,
+			parallel_tool_calls: metadata?.parallelToolCalls ?? true,
 		}
 
 		// Add thinking parameter if reasoning is enabled and model supports it

src/api/providers/cerebras.ts

@@ -146,7 +146,7 @@ export class CerebrasHandler extends BaseProvider implements SingleCompletionHan
 			// Native tool calling support
 			tools: this.convertToolsForOpenAI(metadata?.tools),
 			tool_choice: metadata?.tool_choice,
-			parallel_tool_calls: metadata?.parallelToolCalls ?? false,
+			parallel_tool_calls: metadata?.parallelToolCalls ?? true,
 		}
 
 		try {

src/api/providers/deepinfra.ts

@@ -74,7 +74,7 @@ export class DeepInfraHandler extends RouterProvider implements SingleCompletion
 			prompt_cache_key,
 			tools: this.convertToolsForOpenAI(_metadata?.tools),
 			tool_choice: _metadata?.tool_choice,
-			parallel_tool_calls: _metadata?.parallelToolCalls ?? false,
+			parallel_tool_calls: _metadata?.parallelToolCalls ?? true,
 		} as OpenAI.Chat.Completions.ChatCompletionCreateParamsStreaming
 
 		if (this.supportsTemperature(modelId)) {

src/api/providers/deepseek.ts

@@ -72,7 +72,7 @@ export class DeepSeekHandler extends OpenAiHandler {
 			...(isThinkingModel && { thinking: { type: "enabled" } }),
 			tools: this.convertToolsForOpenAI(metadata?.tools),
 			tool_choice: metadata?.tool_choice,
-			parallel_tool_calls: metadata?.parallelToolCalls ?? false,
+			parallel_tool_calls: metadata?.parallelToolCalls ?? true,
 		}
 
 		// Add max_tokens if needed

src/api/providers/lm-studio.ts

@@ -90,7 +90,7 @@ export class LmStudioHandler extends BaseProvider implements SingleCompletionHan
 				stream: true,
 				tools: this.convertToolsForOpenAI(metadata?.tools),
 				tool_choice: metadata?.tool_choice,
-				parallel_tool_calls: metadata?.parallelToolCalls ?? false,
+				parallel_tool_calls: metadata?.parallelToolCalls ?? true,
 			}
 
 			if (this.options.lmStudioSpeculativeDecodingEnabled && this.options.lmStudioDraftModelId) {

src/api/providers/openai-codex.ts

@@ -321,7 +321,7 @@ export class OpenAiCodexHandler extends BaseProvider implements SingleCompletion
 					}
 				}),
 			tool_choice: metadata?.tool_choice,
-			parallel_tool_calls: metadata?.parallelToolCalls ?? false,
+			parallel_tool_calls: metadata?.parallelToolCalls ?? true,
 		}
 
 		return body

src/api/providers/openai-native.ts

@@ -378,7 +378,7 @@ export class OpenAiNativeHandler extends BaseProvider implements SingleCompletio
 					}
 				}),
 			tool_choice: metadata?.tool_choice,
-			parallel_tool_calls: metadata?.parallelToolCalls ?? false,
+			parallel_tool_calls: metadata?.parallelToolCalls ?? true,
 		}
 
 		// Include text.verbosity only when the model explicitly supports it

src/api/providers/openai.ts

@@ -161,7 +161,7 @@ export class OpenAiHandler extends BaseProvider implements SingleCompletionHandl
 				...(reasoning && reasoning),
 				tools: this.convertToolsForOpenAI(metadata?.tools),
 				tool_choice: metadata?.tool_choice,
-				parallel_tool_calls: metadata?.parallelToolCalls ?? false,
+				parallel_tool_calls: metadata?.parallelToolCalls ?? true,
 			}
 
 			// Add max_tokens if needed
@@ -229,7 +229,7 @@ export class OpenAiHandler extends BaseProvider implements SingleCompletionHandl
 				// Tools are always present (minimum ALWAYS_AVAILABLE_TOOLS)
 				tools: this.convertToolsForOpenAI(metadata?.tools),
 				tool_choice: metadata?.tool_choice,
-				parallel_tool_calls: metadata?.parallelToolCalls ?? false,
+				parallel_tool_calls: metadata?.parallelToolCalls ?? true,
 			}
 
 			// Add max_tokens if needed
@@ -348,7 +348,7 @@ export class OpenAiHandler extends BaseProvider implements SingleCompletionHandl
 				// Tools are always present (minimum ALWAYS_AVAILABLE_TOOLS)
 				tools: this.convertToolsForOpenAI(metadata?.tools),
 				tool_choice: metadata?.tool_choice,
-				parallel_tool_calls: metadata?.parallelToolCalls ?? false,
+				parallel_tool_calls: metadata?.parallelToolCalls ?? true,
 			}
 
 			// O3 family models do not support the deprecated max_tokens parameter
@@ -382,7 +382,7 @@ export class OpenAiHandler extends BaseProvider implements SingleCompletionHandl
 				// Tools are always present (minimum ALWAYS_AVAILABLE_TOOLS)
 				tools: this.convertToolsForOpenAI(metadata?.tools),
 				tool_choice: metadata?.tool_choice,
-				parallel_tool_calls: metadata?.parallelToolCalls ?? false,
+				parallel_tool_calls: metadata?.parallelToolCalls ?? true,
 			}
 
 			// O3 family models do not support the deprecated max_tokens parameter

src/api/providers/qwen-code.ts

@@ -228,7 +228,7 @@ export class QwenCodeHandler extends BaseProvider implements SingleCompletionHan
 			max_completion_tokens: model.info.maxTokens,
 			tools: this.convertToolsForOpenAI(metadata?.tools),
 			tool_choice: metadata?.tool_choice,
-			parallel_tool_calls: metadata?.parallelToolCalls ?? false,
+			parallel_tool_calls: metadata?.parallelToolCalls ?? true,
 		}
 
 		const stream = await this.callApiWithRetry(() => client.chat.completions.create(requestOptions))

src/api/providers/unbound.ts

@@ -121,7 +121,7 @@ export class UnboundHandler extends RouterProvider implements SingleCompletionHa
 			},
 			tools: this.convertToolsForOpenAI(metadata?.tools),
 			tool_choice: metadata?.tool_choice,
-			parallel_tool_calls: metadata?.parallelToolCalls ?? false,
+			parallel_tool_calls: metadata?.parallelToolCalls ?? true,
 		}
 
 		if (this.supportsTemperature(modelId)) {

src/api/providers/vercel-ai-gateway.ts

@@ -63,7 +63,7 @@ export class VercelAiGatewayHandler extends RouterProvider implements SingleComp
 			stream_options: { include_usage: true },
 			tools: this.convertToolsForOpenAI(metadata?.tools),
 			tool_choice: metadata?.tool_choice,
-			parallel_tool_calls: metadata?.parallelToolCalls ?? false,
+			parallel_tool_calls: metadata?.parallelToolCalls ?? true,
 		}
 
 		const completion = await this.client.chat.completions.create(body)

src/api/providers/xai.ts

@@ -68,7 +68,7 @@ export class XAIHandler extends BaseProvider implements SingleCompletionHandler
 			...(reasoning && reasoning),
 			tools: this.convertToolsForOpenAI(metadata?.tools),
 			tool_choice: metadata?.tool_choice,
-			parallel_tool_calls: metadata?.parallelToolCalls ?? false,
+			parallel_tool_calls: metadata?.parallelToolCalls ?? true,
 		}
 
 		let stream

src/api/providers/zai.ts

@@ -103,7 +103,7 @@ export class ZAiHandler extends BaseOpenAiCompatibleProvider<string> {
 			thinking: useReasoning ? { type: "enabled" } : { type: "disabled" },
 			tools: this.convertToolsForOpenAI(metadata?.tools),
 			tool_choice: metadata?.tool_choice,
-			parallel_tool_calls: metadata?.parallelToolCalls ?? false,
+			parallel_tool_calls: metadata?.parallelToolCalls ?? true,
 		}
 
 		return this.client.chat.completions.create(params)