After following the quickstart document to build a real word agent, it aways throws this error:
400 An assistant message with ‘tool_calls’ must be followed by tool messages responding to each ‘tool_call_id’
I think this might be related to that pending bug fix fix(openai): pair reasoning with function_call id; add tests by pawel-twardziak · Pull Request #9082 · langchain-ai/langchainjs · GitHub
I will check it.
Hmmm the example works for me:
{
messages: [
HumanMessage {
"id": "29710469-f7bd-4e30-a0e8-998c35b8173d",
"content": "What's the weather in Tokyo?",
"additional_kwargs": {},
"response_metadata": {}
},
AIMessage {
"id": "msg_01YAxpwQzBjL6MZqqpjHPsQH",
"content": [
{
"type": "tool_use",
"id": "toolu_015E8qECr7A8AkFGHAqUYnnX",
"name": "get_weather",
"input": {
"city": "Tokyo"
}
}
],
"name": "model",
"additional_kwargs": {
"model": "claude-sonnet-4-5-20250929",
"id": "msg_01YAxpwQzBjL6MZqqpjHPsQH",
"type": "message",
"role": "assistant",
"stop_reason": "tool_use",
"stop_sequence": null,
"usage": {
"input_tokens": 606,
"cache_creation_input_tokens": 0,
"cache_read_input_tokens": 0,
"cache_creation": {
"ephemeral_5m_input_tokens": 0,
"ephemeral_1h_input_tokens": 0
},
"output_tokens": 53,
"service_tier": "standard"
}
},
"response_metadata": {
"model": "claude-sonnet-4-5-20250929",
"id": "msg_01YAxpwQzBjL6MZqqpjHPsQH",
"stop_reason": "tool_use",
"stop_sequence": null,
"usage": {
"input_tokens": 606,
"cache_creation_input_tokens": 0,
"cache_read_input_tokens": 0,
"cache_creation": {
"ephemeral_5m_input_tokens": 0,
"ephemeral_1h_input_tokens": 0
},
"output_tokens": 53,
"service_tier": "standard"
},
"type": "message",
"role": "assistant",
"model_provider": "anthropic"
},
"tool_calls": [
{
"name": "get_weather",
"args": {
"city": "Tokyo"
},
"id": "toolu_015E8qECr7A8AkFGHAqUYnnX",
"type": "tool_call"
}
],
"invalid_tool_calls": [],
"usage_metadata": {
"input_tokens": 606,
"output_tokens": 53,
"total_tokens": 659,
"input_token_details": {
"cache_creation": 0,
"cache_read": 0
}
}
},
ToolMessage {
"id": "1ee82833-84f4-4f5e-8482-36e27f09b114",
"content": "It's always sunny in Tokyo!",
"name": "get_weather",
"additional_kwargs": {},
"response_metadata": {},
"tool_call_id": "toolu_015E8qECr7A8AkFGHAqUYnnX"
},
AIMessage {
"id": "msg_01JHisN4bK4L6j129EQ7WCtW",
"content": "The weather in Tokyo is sunny! ☀️",
"name": "model",
"additional_kwargs": {
"model": "claude-sonnet-4-5-20250929",
"id": "msg_01JHisN4bK4L6j129EQ7WCtW",
"type": "message",
"role": "assistant",
"stop_reason": "end_turn",
"stop_sequence": null,
"usage": {
"input_tokens": 678,
"cache_creation_input_tokens": 0,
"cache_read_input_tokens": 0,
"cache_creation": {
"ephemeral_5m_input_tokens": 0,
"ephemeral_1h_input_tokens": 0
},
"output_tokens": 14,
"service_tier": "standard"
}
},
"response_metadata": {
"model": "claude-sonnet-4-5-20250929",
"id": "msg_01JHisN4bK4L6j129EQ7WCtW",
"stop_reason": "end_turn",
"stop_sequence": null,
"usage": {
"input_tokens": 678,
"cache_creation_input_tokens": 0,
"cache_read_input_tokens": 0,
"cache_creation": {
"ephemeral_5m_input_tokens": 0,
"ephemeral_1h_input_tokens": 0
},
"output_tokens": 14,
"service_tier": "standard"
},
"type": "message",
"role": "assistant",
"model_provider": "anthropic"
},
"tool_calls": [],
"invalid_tool_calls": [],
"usage_metadata": {
"input_tokens": 678,
"output_tokens": 14,
"total_tokens": 692,
"input_token_details": {
"cache_creation": 0,
"cache_read": 0
}
}
}
]
}
@DeepenLau have you changed the provider for OpenAI? Or any changes to the code?
For openai:gpt-5 it also works for me:
{
messages: [
HumanMessage {
"id": "9f0c1e16-87d1-429e-8a32-041974071a99",
"content": "What's the weather in Tokyo?",
"additional_kwargs": {},
"response_metadata": {}
},
AIMessage {
"id": "chatcmpl-CT6x0kWc3KJnZHt9Ph1VbOR5ZcCKE",
"content": "",
"name": "model",
"additional_kwargs": {
"tool_calls": [
{
"id": "call_rqQIxuJLO81Jaf69mfwni2V2",
"type": "function",
"function": "[Object]"
}
]
},
"response_metadata": {
"tokenUsage": {
"promptTokens": 141,
"completionTokens": 151,
"totalTokens": 292
},
"finish_reason": "tool_calls",
"model_provider": "openai",
"model_name": "gpt-5-2025-08-07"
},
"tool_calls": [
{
"name": "get_weather",
"args": {
"city": "Tokyo"
},
"type": "tool_call",
"id": "call_rqQIxuJLO81Jaf69mfwni2V2"
}
],
"invalid_tool_calls": [],
"usage_metadata": {
"output_tokens": 151,
"input_tokens": 141,
"total_tokens": 292,
"input_token_details": {
"audio": 0,
"cache_read": 0
},
"output_token_details": {
"audio": 0,
"reasoning": 128
}
}
},
ToolMessage {
"id": "040c7f44-3629-444b-93d2-203265386898",
"content": "It's always sunny in Tokyo!",
"name": "get_weather",
"additional_kwargs": {},
"response_metadata": {},
"tool_call_id": "call_rqQIxuJLO81Jaf69mfwni2V2"
},
AIMessage {
"id": "chatcmpl-CT6x6kv44j5BlB3KNc7EjsoSmFxaj",
"content": "It’s always sunny in Tokyo!",
"name": "model",
"additional_kwargs": {},
"response_metadata": {
"tokenUsage": {
"promptTokens": 176,
"completionTokens": 592,
"totalTokens": 768
},
"finish_reason": "stop",
"model_provider": "openai",
"model_name": "gpt-5-2025-08-07"
},
"tool_calls": [],
"invalid_tool_calls": [],
"usage_metadata": {
"output_tokens": 592,
"input_tokens": 176,
"total_tokens": 768,
"input_token_details": {
"audio": 0,
"cache_read": 0
},
"output_token_details": {
"audio": 0,
"reasoning": 576
}
}
}
]
}
How are you invoking it second time? Have you modified the code? What’s the invocations flow?
I have the same issue with custom graph - running ok for the first time but get this error at the 2nd time.
Had the same error here, same scenario whenever I made a second time call into an already existant graph I would get this error.
downgrading from 1.1.3 to 1.0.0 got it working again, hope it helps
1.1.1 and 1.1.2 both presented the same issue.
@reinvent-heucles Thanks. I’m getting this error with 1.1.0 of langchain-core and langchain. Maybe I downgrade this further to try out…
I could try to fix it, but… it is not happening to me with the latest versions (1.1.x) at all 
this code:
import { createAgent, tool } from "langchain";
import * as z from "zod";
import dotenv from 'dotenv'
import { MemorySaver } from "@langchain/langgraph";
import { initChatModel } from "langchain";
dotenv.config()
const systemPrompt = `You are an expert weather forecaster, who speaks in puns.
You have access to two tools:
- get_weather_for_location: use this to get the weather for a specific location
- get_user_location: use this to get the user's location
If a user asks you for the weather, make sure you know the location. If you can tell from the question that they mean wherever they are, use the get_user_location tool to find their location.`;
const checkpointer = new MemorySaver();
const responseFormat = z.object({
punny_response: z.string(),
weather_conditions: z.string().optional(),
});
const model = await initChatModel(
"deepseek-chat",
{ temperature: 0.5, timeout: 1800, maxTokens: 3200, modelProvider: 'deepseek', maxRetires: 2 }
);
const getWeather = tool(
(input) => `It's always sunny in ${input.city}!`,
{
name: "get_weather_for_location",
description: "Get the weather for a given city",
schema: z.object({
city: z.string().describe("The city to get the weather for"),
}),
}
);
// type AgentRuntime = ToolRuntime<unknown, { user_id: string }>;
const getUserLocation = tool(
(_, config) => {
const { user_id } = config.context;
return user_id === "1" ? "Florida" : "SF";
},
{
name: "get_user_location",
description: "Retrieve user information based on user ID",
}
);
const agent = createAgent({
model,
systemPrompt: systemPrompt,
tools: [getUserLocation, getWeather],
responseFormat,
checkpointer,
});
// `thread_id` is a unique identifier for a given conversation.
const config = {
configurable: { thread_id: "1" },
context: { user_id: "1" },
};
const response = await agent.invoke(
{ messages: [{ role: "user", content: "what is the weather outside?" }] },
config
);
console.log(response.structuredResponse);
// {
// punny_response: "Florida is still having a 'sun-derful' day ...",
// weather_conditions: "It's always sunny in Florida!"
// }
// Note that we can continue the conversation using the same `thread_id`.
const thankYouResponse = await agent.invoke(
{ messages: [{ role: "user", content: "thank you!" }] },
config
);
console.log(thankYouResponse.structuredResponse);
// {
// punny_response: "You're 'thund-erfully' welcome! ...",
// weather_conditions: undefined
// }
has successfully produced 5 answers in a row like this:
{
punny_response: `Well, it looks like Florida is living up to its sunny reputation! I'd say the weather is absolutely "sun-sational" - you could say it's a real "bright" spot on the map! Don't forget your sunscreen, or you might get "burned" by that Florida sunshine!`,
weather_conditions: 'Sunny'
}
{
punny_response: `You're very welcome! I'm always "weather" to help - it's my "fore-cast" in life! If you need any more weather "pun-dits", just give me a "cloud" shout!`
}
Have you tried to remove all the dependencies phisically and install them from scratch?