Browser Tool + Memory: orphaned toolUse causing ValidationException after first message

[very99]

Bug Description

When using AgentCoreBrowser (Browser Tool) together with AgentCoreMemorySessionManager (AgentCore Memory), the agent fails with a ValidationException after the first successful message in the same session.

Environment

• Strands Agents: Latest version
• bedrock-agentcore: Latest version
• strands-agents-tools: Latest version
• Region: us-east-1
• Runtime: AgentCore Runtime (JWT authentication)

Problem Description

Error Message

Session message history has an orphaned toolUse with no toolResult. 
Adding toolResult content blocks to create valid conversation.

ValidationException: The number of toolResult blocks at messages.27.content 
exceeds the number of toolUse blocks of previous turn.

Behavior

• ✅ First message in session: Works correctly
• ❌ Subsequent messages in same session: Fails with ValidationException
• ✅ Without Browser Tool (Gateway tools only): Works correctly
• ❌ Without Gateway (Browser Tool + Memory only): Still fails

Code Example

from strands import Agent
from strands_tools.browser import AgentCoreBrowser
from bedrock_agentcore.memory.integrations.strands.config import AgentCoreMemoryConfig
from bedrock_agentcore.memory.integrations.strands.session_manager import AgentCoreMemorySessionManager

# Memory configuration
memory_config = AgentCoreMemoryConfig(
    memory_id=MEMORY_ID,
    session_id=session_id,
    actor_id=actor_id
)

session_manager = AgentCoreMemorySessionManager(
    agentcore_memory_config=memory_config,
    region_name="us-east-1"
)

# Browser Tool
browser_tool = AgentCoreBrowser(region="us-east-1")

# Agent with Browser Tool + Memory
agent = Agent(
    model="us.anthropic.claude-sonnet-4-5-20250929-v1:0",
    system_prompt="You are a helpful assistant.",
    session_manager=session_manager,
    tools=[browser_tool.browser]
)

# First call: SUCCESS ✅
result1 = agent("Hello")

# Second call in same session: FAILS ❌
result2 = agent("How are you?")

Root Cause Analysis

Based on similar issues found in the community:

1. Anthropic Claude Code Issue #13964: "orphaned tool_result blocks after context truncation"

   • When conversation history is truncated/summarized, tool_use blocks are removed
   • But corresponding tool_result blocks are preserved
   • This creates orphaned tool results

2. browser-use Issue [BUG] Claude 4 models stop at 8192 output tokens #710: Similar ValidationException with Bedrock Converse API

Hypothesis

The issue appears to be related to how AgentCore Memory saves conversation history when Browser Tool is used:

1. Browser Tool creates tool_use blocks
2. Memory saves the conversation history
3. On subsequent requests, Memory loads the history
4. Something in the Memory/Strands integration causes tool_use/tool_result mismatch
5. Strands detects "orphaned toolUse" and tries to auto-repair
6. The auto-repair logic adds extra toolResult blocks
7. Bedrock API rejects the request due to count mismatch

Logs

2026-02-02T10:12:24.909000+00:00 Session message history has an orphaned toolUse with no toolResult. Adding toolResult content blocks to create valid conversation.

2026-02-02T10:12:25.876000+00:00 botocore.errorfactory.ValidationException: An error occurred (ValidationException) when calling the ConverseStream operation: The number of toolResult blocks at messages.27.content exceeds the number of toolUse blocks of previous turn.

Questions

1. Is Browser Tool + Memory a supported combination?
2. Is there official documentation for using Browser Tool with Memory?
3. What is the "orphaned toolUse" auto-repair mechanism? Is it documented?
4. Should Memory handle Browser Tool's conversation history differently?

Expected Behavior

Browser Tool and Memory should work together seamlessly without ValidationException errors across multiple messages in the same session.

Actual Behavior

After the first successful message, subsequent messages in the same session fail with ValidationException about toolResult/toolUse count mismatch.

Related Issues

• Anthropic Claude Code #13964: orphaned tool_result blocks after context truncation
• browser-use [BUG] Claude 4 models stop at 8192 output tokens #710: Browser-use fails with Bedrock Converse API

Suggested Fix

Similar to the fix suggested in Anthropic's issue, add a sanitization function that:

1. Collects all tool_use_id values from tool_use blocks
2. Removes any tool_result blocks whose tool_use_id doesn't have a matching tool_use block
3. Runs this sanitization before every API call

This should be done either in:

• Strands Agents' conversation history management
• Or in the AgentCore Memory integration layer

[cagataycali]

Great bug report with detailed root cause analysis @very99!

Technical Analysis

I've examined the _fix_broken_tool_use method in src/strands/session/repository_session_manager.py (lines 165-228). Here's what I found:

Current Logic

1. Line 185-193: Removes orphaned toolResult at the START of messages (handles truncation case)
2. Line 200-227: For each toolUse message, finds missing toolResults and ADDS them

The Problem

The sanitization adds toolResult blocks to messages[index + 1] whenever there's a mismatch:

missing_tool_use_ids = list(set(tool_use_ids) - set(tool_result_ids))
if missing_tool_use_ids:
    missing_content_blocks = generate_missing_tool_result_content(missing_tool_use_ids)
    messages[index + 1]["content"].extend(missing_content_blocks)  # Line 225

But this assumes messages[index + 1] ONLY contains toolResults for messages[index]'s toolUses. If AgentCore Memory stores conversations differently (e.g., interleaved tool calls, or conversation context from browser operations), this assumption breaks.

Why "exceeds the number of toolUse blocks"

The error occurs because:

1. Message N has X toolUse blocks
2. Message N+1 already has toolResult blocks (possibly from different operations)
3. Sanitization adds MORE toolResult blocks to N+1
4. Total toolResult count > X, causing Bedrock API rejection

Missing Logic

The code handles:

• ✅ Orphaned toolResult at START (removes)
• ✅ Missing toolResult after toolUse (adds)

But doesn't handle:

• ❌ Orphaned toolResult blocks in the MIDDLE of conversation
• ❌ toolResult blocks that don't have matching toolUse in PREVIOUS message specifically

Suggested Fix

Add a complementary sanitization step that removes orphaned toolResult blocks:

# After line 196, add:
for index, message in enumerate(messages):
    if index > 0 and message["role"] == "user":
        # Get tool_use_ids from previous assistant message
        prev_msg = messages[index - 1]
        if prev_msg["role"] == "assistant":
            valid_tool_use_ids = {
                c["toolUse"]["toolUseId"] 
                for c in prev_msg["content"] 
                if "toolUse" in c
            }
            # Filter toolResult blocks to only those with matching toolUse
            message["content"] = [
                c for c in message["content"]
                if "toolResult" not in c or c["toolResult"]["toolUseId"] in valid_tool_use_ids
            ]

This would ensure the 1:1 correspondence between toolUse and toolResult that Bedrock requires.

---

🤖 AI agent response from the Strands team. Strands Agents. Feedback welcome!

[very99]

🔍 Additional Research: This is a Widespread Known Issue

After further investigation, I found that this "orphaned toolUse/toolResult" error is a widespread known issue affecting multiple projects and frameworks, not just Strands Agents.

📋 Related Issues Across Multiple Projects

1. Anthropic Claude Code #14416 - "Context compaction breaks tool_use/tool_result pairing"

   • 🔗 Context compaction breaks tool_use/tool_result pairing causing API errors anthropics/claude-code#14416
   • Root Cause: During automatic context compaction in long sessions, the system sometimes removes tool_use blocks while keeping corresponding tool_result blocks, breaking the required API pairing and corrupting the conversation.

2. Anthropic Claude Code #14173 - "orphaned tool_result in messages array despite valid conversation history"

   • 🔗 Compaction error: orphaned tool_result in messages array despite valid conversation history anthropics/claude-code#14173
   • Similar orphaned tool_result blocks issue

3. Anthropic Claude Code #12894 - Related bug report

   • Referenced by multiple developers as a known issue

4. Pydantic AI #2327 - "Bedrock Agent Error: The number of toolResult blocks exceeds..."

   • 🔗 Bedrock Agent Error:The number of toolResult blocks exceeds the number of toolUse blocks of previous turn pydantic/pydantic-ai#2327
   • Exact same error in Pydantic AI framework

5. OpenHands #4912 - "BedrockException: The number of toolResult blocks at messages.2.content exceeds..."

   • 🔗 [Bug]: BedrockException: "The number of toolResult blocks at messages.2.content exceeds the number of toolUse blocks of previous turn.". OpenHands/OpenHands#4912
   • Exact same error in OpenHands project

6. LiteLLM #7099 - "The wrong number of toolUse blocks when using AWS Bedrock"

   • 🔗 [Bug]: The wrong number of toolUse blocks when using AWS Bedrock BerriAI/litellm#7099
   • Exact same error in LiteLLM framework

🎯 Root Cause (from Anthropic Issue #14416)

"During automatic context compaction in long sessions, Claude Code sometimes removes a tool_use block while keeping its corresponding tool_result block. This breaks the required API pairing and causes the conversation to become corrupted."

💡 Why Browser Tool + Memory Triggers This

In my case, the combination of Browser Tool + Memory likely triggers this issue because:

1. Browser Tool generates many tool calls (navigation, clicking, reading, etc.)
2. Memory persists the conversation history across requests
3. As the session grows longer, some context compaction/truncation mechanism is triggered
4. During compaction, tool_use blocks are removed but tool_result blocks remain
5. This creates orphaned tool results, causing the ValidationException

This matches the pattern described in Anthropic Issue #14416 exactly.

📚 Community Workarounds

From https://calvin.my/posts/api-error-claude-block-must-have-a-corresponding-tool_use-block-in-the-previous-message:

Temporary workarounds include:

1. End the current conversation and start a new one
2. Use /compact command to wipe conversation history while maintaining summarized context
3. Use /rewind command to go back 1-2 steps

✅ Conclusion

This is not a Strands-specific issue but rather a fundamental problem with context management in tool-using conversations that affects:

• Anthropic Claude Code
• Pydantic AI
• OpenHands
• LiteLLM
• Strands Agents (this issue)

The issue needs to be addressed at the conversation history management layer to properly handle tool_use/tool_result pairing during context compaction/truncation.

🔧 Suggested Fix (from community)

A sanitization function should be added that:

1. Collects all tool_use_id values from tool_use blocks
2. Removes any tool_result blocks whose tool_use_id doesn't have a matching tool_use block
3. Runs this sanitization before every API call and after loading from memory

This should prevent orphaned tool blocks from reaching the Bedrock API.