springBeanToolCallbackResolver want to create a bean which is not existed

[daicidemeigui]

I only used the MCP client to create methods and did not use Bean to construct them. However, when calling a method that does not exist in the MCP client, I will use the resolve method of the SpringBeanToolFHIR Parser created in ToolCallingAutoConfiguration to create the Bean. At this point, the program will directly report an error:
java.lang.IllegalArgumentException: Functional bean with name mcp_client_filesystem_delete_file does not exist in the context.

[tzolov]

Hey @daicidemeigui , can you please elaborate on this issue. Is it possible to reproduce it?

[yvannh]

Hey, I can tell that I got the same issue yesterday and I dont think it can be reproduced (maybe just in my case though), so I'll be just describing the architecture and what went wrong. I have two servers exposing tools (one is for translation - Translation MCP - and one for a database operations - Movie MCP, also acting as a client). When the Movie MCP client was asking for some translation, I got the issue on one of them, it did well but at this step no. The error is that the tool was misnamed with some random uppercase character (from tool calling trace : Executing tool call: spring_ai_mcp_client_translator_mCP_translate, instead of just spring_ai_mcp_client_translator_mcp_translate, mcp with all lower case). When I reran, the issue happened on the same translation I don't know why, there was no explanation on that. To "fix" the issue, I just reloaded the translation MCP server, the issue disappeared, but in my opinion it can reappear again. If you need more details, do not hesitate.

Edit : the issue reappeared... and on the same translation it was previously trying, I don't get it why. Here are the logs, maybe it says something :

2025-04-10T14:29:54.668+02:00  INFO 7472 --- [movie-search-mcp] [           main] f.y.m.services.MovieFactory              : Translation in progress...
2025-04-10T14:29:56.314+02:00 DEBUG 7472 --- [movie-search-mcp] [           main] o.s.a.m.tool.DefaultToolCallingManager   : Executing tool call: spring_ai_mcp_client_translator_mCP_translate
2025-04-10T14:29:56.314+02:00 DEBUG 7472 --- [movie-search-mcp] [           main] o.s.a.t.r.StaticToolCallbackResolver     : ToolCallback resolution attempt from static registry
2025-04-10T14:29:56.314+02:00 DEBUG 7472 --- [movie-search-mcp] [           main] o.s.a.t.r.SpringBeanToolCallbackResolver : ToolCallback resolution attempt from Spring application context
Exception in thread "main" java.lang.IllegalArgumentException: Functional bean with name spring_ai_mcp_client_translator_mCP_translate does not exist in the context.
	at org.springframework.ai.tool.resolution.TypeResolverHelper.getBeanDefinition(TypeResolverHelper.java:146)
	at org.springframework.ai.tool.resolution.TypeResolverHelper.resolveBeanType(TypeResolverHelper.java:124)
	at org.springframework.ai.tool.resolution.SpringBeanToolCallbackResolver.resolve(SpringBeanToolCallbackResolver.java:91)
	at org.springframework.ai.tool.resolution.SpringBeanToolCallbackResolver.resolve(SpringBeanToolCallbackResolver.java:59)
	at org.springframework.ai.tool.resolution.DelegatingToolCallbackResolver.resolve(DelegatingToolCallbackResolver.java:48)
	at org.springframework.ai.model.tool.DefaultToolCallingManager.lambda$executeToolCall$4(DefaultToolCallingManager.java:206)
	at java.base/java.util.Optional.orElseGet(Optional.java:364)
	at org.springframework.ai.model.tool.DefaultToolCallingManager.executeToolCall(DefaultToolCallingManager.java:206)
	at org.springframework.ai.model.tool.DefaultToolCallingManager.executeToolCalls(DefaultToolCallingManager.java:139)
	at org.springframework.ai.ollama.OllamaChatModel.internalCall(OllamaChatModel.java:250)
	at org.springframework.ai.ollama.OllamaChatModel.call(OllamaChatModel.java:201)
	at org.springframework.ai.chat.client.DefaultChatClient$DefaultChatClientRequestSpec$1.aroundCall(DefaultChatClient.java:680)
	at org.springframework.ai.chat.client.advisor.DefaultAroundAdvisorChain.lambda$nextAroundCall$1(DefaultAroundAdvisorChain.java:98)
	at io.micrometer.observation.Observation.observe(Observation.java:564)
	at org.springframework.ai.chat.client.advisor.DefaultAroundAdvisorChain.nextAroundCall(DefaultAroundAdvisorChain.java:98)
	at org.springframework.ai.chat.client.DefaultChatClient$DefaultCallResponseSpec.doGetChatResponse(DefaultChatClient.java:493)
	at org.springframework.ai.chat.client.DefaultChatClient$DefaultCallResponseSpec.lambda$doGetObservableChatResponse$1(DefaultChatClient.java:482)
	at io.micrometer.observation.Observation.observe(Observation.java:564)
	at org.springframework.ai.chat.client.DefaultChatClient$DefaultCallResponseSpec.doGetObservableChatResponse(DefaultChatClient.java:482)
	at org.springframework.ai.chat.client.DefaultChatClient$DefaultCallResponseSpec.doGetChatResponse(DefaultChatClient.java:466)
	at org.springframework.ai.chat.client.DefaultChatClient$DefaultCallResponseSpec.content(DefaultChatClient.java:516)

[daicidemeigui]

Hey @daicidemeigui , can you please elaborate on this issue. Is it possible to reproduce it?

Maybe it's a model illusion, because I deployed a low parameter model locally, and he may have created a non-existent tool based on the provided tool, but I think this is also a bug

[daicidemeigui]

Hey @daicidemeigui , can you please elaborate on this issue. Is it possible to reproduce it?

The scene at that time was like this:
My question is' help me to delete a file named 'test. txt' under'D: '/ 006765/Desktop/claude mcp examples main '", then using the following two open-source MCP services, configured as follows:
{
"mcpServers": {
"playwright": {
"command": "node",
"args": [
"-e",
"require('child_process').execSync('npx -y @executeautomation/playwright-mcp-server', {stdio: 'inherit'})"
]
},
"filesystem": {
"command": "node",
"args": [
"-e",
"require('child_process').execSync('npx -y @modelcontextprotocol/server-filesystem .', {stdio: 'inherit'})"
]
}
}
}
I only used the Tool provided by MCP service to the model, and did not write any new Tool in the code. Due to the illusion generated by the model, it called a method that was not provided in the MCP service, but attempted to create a Bean for that method in the SpringAI framework, resulting in this issue

[daicidemeigui]

Hey @daicidemeigui , can you please elaborate on this issue. Is it possible to reproduce it?

This is the code which is i use to chat:
@resource
private OllamaChatModel ollamaChatModel;

    @Resource
    private AsyncMcpToolCallbackProvider toolCallbackProvider;

    @PostMapping("/chat")
    public ResponseEntity<ChatResponse> call(@RequestBody ChatRequest input) {
        try {
            ChatClient chatClient = ChatClient.builder(ollamaChatModel)
                    .defaultTools(toolCallbackProvider.getToolCallbacks())
                    .build();
            String content = chatClient.prompt(input.getMessage()).call().content();
            return ResponseEntity.ok(new ChatResponse(content));
        } catch (Exception e) {
            e.printStackTrace();
            return ResponseEntity.ok(new ChatResponse("ERROR: " + e.getMessage()));
        }

    }

The error is generated in SpringBeanToolCallbackResolver.class.The error occurred when the resolve method in this class called the resolveBeanType method because it attempted to generate a bean

[Ga-Ol-St]

I also hit this issue,
i'm using M7,
have combination of few MCP clients + additional Tools in classes, using usual annotation "tool"
but i'm using "User-Controlled Tool Execution" approach.

I traced it to
org.springframework.ai.model.tool.autoconfigure.ToolCallingAutoConfiguration#toolCallbackResolver

it's registering SpringBeanToolCallbackResolver as a delegate
and if you anywhere try to run tool which is not registered - it will go to springBeanToolCallbackResolver and die because can't find bean with same name as Tool name.

in my case it's happening here:
org.springframework.ai.model.tool.DefaultToolCallingManager#resolveToolDefinitions
when i registering only toolNames (not callbacks) via
ChatOptions toolOptions = ToolCallingChatOptions.builder().toolNames(

Why toolnames - there is an another issue which don't allow me to use MCP callbacks in "User-Controlled Tool Execution" (and probably other)

issue:"New MCP tools naming schema is creating mismatch with prompt.options in org.springframework.ai.model.tool.DefaultToolCallingManager#executeToolCall"

I did manage to trace everything to org.springframework.ai.aot.ToolBeanRegistrationAotProcessor - it's not starting and did not registered all my Tool annotations, so no automatic tool registration, why it's not started - still digging.

P.S. All this issues appeared after i switched to M7.

Java.lang.IllegalArgumentException: Functional bean with name getWrapLog does not exist in the context.
	at org.springframework.ai.tool.resolution.TypeResolverHelper.getBeanDefinition(TypeResolverHelper.java:146) ~[spring-ai-model-1.0.0-M7.jar:1.0.0-M7]
	at org.springframework.ai.tool.resolution.TypeResolverHelper.resolveBeanType(TypeResolverHelper.java:124) ~[spring-ai-model-1.0.0-M7.jar:1.0.0-M7]
	at org.springframework.ai.tool.resolution.SpringBeanToolCallbackResolver.resolve(SpringBeanToolCallbackResolver.java:91) ~[spring-ai-model-1.0.0-M7.jar:1.0.0-M7]
	at org.springframework.ai.tool.resolution.SpringBeanToolCallbackResolver.resolve(SpringBeanToolCallbackResolver.java:59) ~[spring-ai-model-1.0.0-M7.jar:1.0.0-M7]
	at org.springframework.ai.tool.resolution.DelegatingToolCallbackResolver.resolve(DelegatingToolCallbackResolver.java:48) ~[spring-ai-model-1.0.0-M7.jar:1.0.0-M7]
	at org.springframework.ai.model.tool.DefaultToolCallingManager.resolveToolDefinitions(DefaultToolCallingManager.java:100) ~[spring-ai-model-1.0.0-M7.jar:1.0.0-M7]
	at org.springframework.ai.openai.OpenAiChatModel.createRequest(OpenAiChatModel.java:606) ~[spring-ai-openai-1.0.0-M7.jar:1.0.0-M7]
	at org.springframework.ai.openai.OpenAiChatModel.internalCall(OpenAiChatModel.java:185) ~[spring-ai-openai-1.0.0-M7.jar:1.0.0-M7]
	at org.springframework.ai.openai.OpenAiChatModel.call(OpenAiChatModel.java:180) ~[spring-ai-openai-1.0.0-M7.jar:1.0.0-M7]
	at org.springframework.ai.chat.client.DefaultChatClient$DefaultChatClientRequestSpec$1.aroundCall(DefaultChatClient.java:680) ~[spring-ai-client-chat-1.0.0-M7.jar:1.0.0-M7]
	at org.springframework.ai.chat.client.advisor.DefaultAroundAdvisorChain.lambda$nextAroundCall$1(DefaultAroundAdvisorChain.java:98) ~[spring-ai-client-chat-1.0.0-M7.jar:1.0.0-M7]
	at io.micrometer.observation.Observation.observe(Observation.java:564) ~[micrometer-observation-1.14.5.jar:1.14.5]
	at org.springframework.ai.chat.client.advisor.DefaultAroundAdvisorChain.nextAroundCall(DefaultAroundAdvisorChain.java:98) ~[spring-ai-client-chat-1.0.0-M7.jar:1.0.0-M7]
	at org.springframework.ai.chat.client.advisor.SimpleLoggerAdvisor.aroundCall(SimpleLoggerAdvisor.java:99) ~[spring-ai-client-chat-1.0.0-M7.jar:1.0.0-M7]
	at org.springframework.ai.chat.client.advisor.DefaultAroundAdvisorChain.lambda$nextAroundCall$1(DefaultAroundAdvisorChain.java:98) ~[spring-ai-client-chat-1.0.0-M7.jar:1.0.0-M7]
	at io.micrometer.observation.Observation.observe(Observation.java:564) ~[micrometer-observation-1.14.5.jar:1.14.5]
	at org.springframework.ai.chat.client.advisor.DefaultAroundAdvisorChain.nextAroundCall(DefaultAroundAdvisorChain.java:98) ~[spring-ai-client-chat-1.0.0-M7.jar:1.0.0-M7]
	at org.springframework.ai.chat.client.advisor.MessageChatMemoryAdvisor.aroundCall(MessageChatMemoryAdvisor.java:64) ~[spring-ai-client-chat-1.0.0-M7.jar:1.0.0-M7]
	at org.springframework.ai.chat.client.advisor.DefaultAroundAdvisorChain.lambda$nextAroundCall$1(DefaultAroundAdvisorChain.java:98) ~[spring-ai-client-chat-1.0.0-M7.jar:1.0.0-M7]
	at io.micrometer.observation.Observation.observe(Observation.java:564) ~[micrometer-observation-1.14.5.jar:1.14.5]
	at org.springframework.ai.chat.client.advisor.DefaultAroundAdvisorChain.nextAroundCall(DefaultAroundAdvisorChain.java:98) ~[spring-ai-client-chat-1.0.0-M7.jar:1.0.0-M7]
	at org.springframework.ai.chat.client.DefaultChatClient$DefaultCallResponseSpec.doGetChatResponse(DefaultChatClient.java:493) ~[spring-ai-client-chat-1.0.0-M7.jar:1.0.0-M7]
	at org.springframework.ai.chat.client.DefaultChatClient$DefaultCallResponseSpec.lambda$doGetObservableChatResponse$1(DefaultChatClient.java:482) ~[spring-ai-client-chat-1.0.0-M7.jar:1.0.0-M7]
	at io.micrometer.observation.Observation.observe(Observation.java:564) ~[micrometer-observation-1.14.5.jar:1.14.5]
	at org.springframework.ai.chat.client.DefaultChatClient$DefaultCallResponseSpec.doGetObservableChatResponse(DefaultChatClient.java:482) ~[spring-ai-client-chat-1.0.0-M7.jar:1.0.0-M7]
	at org.springframework.ai.chat.client.DefaultChatClient$DefaultCallResponseSpec.doGetChatResponse(DefaultChatClient.java:466) ~[spring-ai-client-chat-1.0.0-M7.jar:1.0.0-M7]
	at org.springframework.ai.chat.client.DefaultChatClient$DefaultCallResponseSpec.chatResponse(DefaultChatClient.java:510) ~[spring-ai-client-chat-1.0.0-M7.jar:1.0.0-M7]
	at com.example.demo.MyController.processRequest(MyController.java:208) 

[tzolov]

@Ga-Ol-St , @daicidemeigui , @yvannh thank you for the feedback , please that M8 has unfortunate breaking change when registering MCP provider. Instead of .defaultTools(...), .tools(...) you should use .defaultToolCallbacks(...) and .toolCallbacks(...).

@Ga-Ol-St are you building native image? I'm puzzled by the ToolBeanRegistrationAotProcessor issue?

[Ga-Ol-St]

@tzolov Hi, no I don't build native image.
In previous version it was enough to annotate some method with Tool annotation and then I was able to use injected Component/Bean class everywhere to reference tools, after switching to M7 - that stopped working and I did hit issue with springBeanToolCallbackResolver so I start searching who is responsible for Tool annotation processing and came to ToolBeanRegistrationAotProcessor
eventually I did manage to resolve both issues by manually injecting my tool class by creating ToolCallBackProvider which will be picked up by ToolCallingAutoConfiguration to ToolCallbackResolver and it's will be used by ToolCallingManager

So the final fix was to manually create ToolCallbackProvider which represent my tool class

@Component
public class MyConfigurationBean {

    @Bean
    ToolCallbackProvider toolCallbackProvider(MyAITools myAITools) {
        return MethodToolCallbackProvider.builder()
                .toolObjects(myAITools)
                .build();

    }

}

[tzolov]

The #3019 should improve the consistency a bit. The individual ToolCallbackResolver implementations should return null and not throwing exceptions to give a chance for the other resolver to kick in.
Now if the tool name is not found you should not see exceptions like :

Java.lang.IllegalArgumentException: Functional bean with name getWrapLog does not exist in the context.
	at org.springframework.ai.tool.resolution.TypeResolverHelper.getBeanDefinition(TypeResolverHelper.java:146) ~[spring-ai-model-1.0.0-M7.jar:1.0.0-M7]
	at org.springframework.ai.tool.resolution.TypeResolverHelper.resolveBeanType(TypeResolverHelper.java:124) ~[spring-ai-model-1.0.0-M7.jar:1.0.0-M7]
	at org.springframework.ai.tool.resolution.SpringBeanToolCallbackResolver.resolve(SpringBeanToolCallbackResolver.java:91) ~[spring-ai-model-1.0.0-M7.jar:1.0.0-M7]
	at org.springframework.ai.tool.resolution.SpringBeanToolCallbackResolver.resolve(SpringBeanToolCallbackResolver.java:59) ~[spring-ai-model-1.0.0-M7.jar:1.0.0-M7]
	at org.springframework.ai.tool.resolution.DelegatingToolCallbackResolver.resolve(DelegatingToo

instead if none of the chained toolcallback resolvers can find the tool name then the DefaultToolCallingManager will trow the exception

[tzolov]

We can't fix the model's tool-name hallucination issue. If the problem is only letter casing, perhaps we could enforce lowercase by default for all MCP tool names?