diff --git a/astrbot/core/agent/context/compressor.py b/astrbot/core/agent/context/compressor.py
index 759604dd93..00d9c61d61 100644
--- a/astrbot/core/agent/context/compressor.py
+++ b/astrbot/core/agent/context/compressor.py
@@ -130,6 +130,7 @@ def __init__(
         instruction_text: str | None = None,
         compression_threshold: float = 0.82,
         token_counter: TokenCounter | None = None,
+        conversation_id: str | None = None,
     ) -> None:
         """Initialize the LLM summary compressor.
 
@@ -139,8 +140,11 @@ def __init__(
                 exact context. Clamped to 0-0.3.
             instruction_text: Custom instruction for summary generation.
             compression_threshold: The compression trigger threshold (default: 0.82).
+            token_counter: Token counter used to preserve recent context.
+            conversation_id: Conversation UUID for the summary request.
         """
         self.provider = provider
+        self.conversation_id = conversation_id
         self.keep_recent_ratio = min(max(float(keep_recent_ratio), 0.0), 0.3)
         self.compression_threshold = compression_threshold
         self.token_counter = token_counter or EstimateTokenCounter()
@@ -275,6 +279,7 @@ async def __call__(self, messages: list[Message]) -> list[Message]:
         try:
             response = await self.provider.text_chat(
                 contexts=sanitized_summary_contexts,
+                conversation_id=self.conversation_id,
             )
             summary_content = (response.completion_text or "").strip()
         except Exception as e:
diff --git a/astrbot/core/agent/context/manager.py b/astrbot/core/agent/context/manager.py
index 1a11ebff96..66b365a2dd 100644
--- a/astrbot/core/agent/context/manager.py
+++ b/astrbot/core/agent/context/manager.py
@@ -13,6 +13,7 @@ class ContextManager:
     def __init__(
         self,
         config: ContextConfig,
+        conversation_id: str | None = None,
     ) -> None:
         """Initialize the context manager.
 
@@ -22,6 +23,7 @@ def __init__(
 
         Args:
             config: The context configuration.
+            conversation_id: Conversation UUID forwarded to summary requests.
         """
         self.config = config
 
@@ -36,6 +38,7 @@ def __init__(
                 keep_recent_ratio=config.llm_compress_keep_recent_ratio,
                 instruction_text=config.llm_compress_instruction,
                 token_counter=self.token_counter,
+                conversation_id=conversation_id,
             )
         else:
             self.compressor = TruncateByTurnsCompressor(
diff --git a/astrbot/core/agent/runners/tool_loop_agent_runner.py b/astrbot/core/agent/runners/tool_loop_agent_runner.py
index c9787ed6f0..b5d9c3c73c 100644
--- a/astrbot/core/agent/runners/tool_loop_agent_runner.py
+++ b/astrbot/core/agent/runners/tool_loop_agent_runner.py
@@ -229,9 +229,18 @@ async def reset(
         request_max_retries: int | None = None,
         tool_result_overflow_dir: str | None = None,
         read_tool: FunctionTool | None = None,
+        # stable identity for plugin-managed conversations when
+        # request.conversation is None (e.g. Context.tool_loop_agent)
+        conversation_id: str | None = None,
         **kwargs: T.Any,
     ) -> None:
         self.req = request
+        # Transient agents need one identity across tool calls and summary requests.
+        self._conversation_id = (
+            request.conversation.cid
+            if request.conversation is not None
+            else (conversation_id or uuid.uuid4().hex)
+        )
         self.streaming = streaming
         self.enforce_max_turns = enforce_max_turns
         self.llm_compress_instruction = llm_compress_instruction
@@ -257,7 +266,8 @@ async def reset(
             custom_compressor=self.custom_compressor,
         )
         self.request_context_manager = ContextManager(
-            self.request_context_manager_config
+            self.request_context_manager_config,
+            conversation_id=self._conversation_id,
         )
 
         self.provider = provider
@@ -505,6 +515,7 @@ async def _iter_llm_responses(
             "contexts": self._sanitize_contexts_for_provider(self.run_context.messages),
             "func_tool": self._func_tool_for_provider(),
             "session_id": self.req.session_id,
+            "conversation_id": self._conversation_id,
             "extra_user_content_parts": self.req.extra_user_content_parts,  # list[ContentPart]
             "abort_signal": self._abort_signal,
             "request_max_retries": self.request_max_retries,
@@ -1447,6 +1458,7 @@ async def _resolve_tool_exec(
                         func_tool=param_subset,
                         model=self.req.model,
                         session_id=self.req.session_id,
+                        conversation_id=self._conversation_id,
                         extra_user_content_parts=self.req.extra_user_content_parts,
                         # tool_choice="required",
                         abort_signal=self._abort_signal,
@@ -1479,6 +1491,7 @@ async def _resolve_tool_exec(
                             func_tool=param_subset,
                             model=self.req.model,
                             session_id=self.req.session_id,
+                            conversation_id=self._conversation_id,
                             extra_user_content_parts=self.req.extra_user_content_parts,
                             # tool_choice="required",
                             abort_signal=self._abort_signal,
diff --git a/astrbot/core/astr_main_agent.py b/astrbot/core/astr_main_agent.py
index ce86e2c18e..1fb9681aef 100644
--- a/astrbot/core/astr_main_agent.py
+++ b/astrbot/core/astr_main_agent.py
@@ -692,6 +692,7 @@ async def _request_img_caption(
     cfg: dict,
     image_urls: list[str],
     plugin_context: Context,
+    conversation_id: str | None = None,
 ) -> str:
     prov = plugin_context.get_provider_by_id(provider_id)
     if prov is None:
@@ -711,6 +712,7 @@ async def _request_img_caption(
     llm_resp = await prov.text_chat(
         prompt=img_cap_prompt,
         image_urls=image_urls,
+        conversation_id=conversation_id,
     )
     return llm_resp.completion_text
 
@@ -728,6 +730,7 @@ async def _ensure_img_caption(
             cfg,
             req.image_urls,
             plugin_context,
+            conversation_id=req.conversation.cid if req.conversation else None,
         )
         if caption:
             req.extra_user_content_parts.append(
@@ -881,6 +884,9 @@ async def _process_quote_message(
                     llm_resp = await prov.text_chat(
                         prompt="Please describe the image content.",
                         image_urls=[image_ref],
+                        conversation_id=req.conversation.cid
+                        if req.conversation
+                        else None,
                     )
                     if llm_resp.completion_text:
                         content_parts.append(
@@ -1019,6 +1025,7 @@ async def _handle_webchat(
 
     try:
         llm_resp = await prov.text_chat(
+            conversation_id=req.conversation.cid if req.conversation else None,
             system_prompt=(
                 "You are a conversation title generator. "
                 "Generate a concise title in the same language as the user’s input, "
diff --git a/astrbot/core/config/default.py b/astrbot/core/config/default.py
index 9cc95a13ed..aaf1f19ffb 100644
--- a/astrbot/core/config/default.py
+++ b/astrbot/core/config/default.py
@@ -1315,6 +1315,42 @@
                         "proxy": "",
                         "custom_headers": {},
                     },
+                    "OpenCode Go Chat Completions": {
+                        "id": "opencode-go",
+                        "provider": "opencode-go",
+                        "type": "opencode_go_chat_completion",
+                        "provider_type": "chat_completion",
+                        "enable": True,
+                        "key": [],
+                        "api_base": "https://opencode.ai/zen/go/v1",
+                        "timeout": 120,
+                        "proxy": "",
+                        "custom_headers": {},
+                    },
+                    "OpenCode Go Responses": {
+                        "id": "opencode-go-responses",
+                        "provider": "opencode-go",
+                        "type": "opencode_go_responses",
+                        "provider_type": "chat_completion",
+                        "enable": True,
+                        "key": [],
+                        "api_base": "https://opencode.ai/zen/go/v1",
+                        "timeout": 120,
+                        "proxy": "",
+                        "custom_headers": {},
+                    },
+                    "OpenCode Go Messages": {
+                        "id": "opencode-go-messages",
+                        "provider": "opencode-go",
+                        "type": "opencode_go_messages",
+                        "provider_type": "chat_completion",
+                        "enable": True,
+                        "key": [],
+                        "api_base": "https://opencode.ai/zen/go/v1",
+                        "timeout": 120,
+                        "proxy": "",
+                        "custom_headers": {},
+                    },
                     "Google Gemini": {
                         "id": "google_gemini",
                         "provider": "google",
diff --git a/astrbot/core/provider/manager.py b/astrbot/core/provider/manager.py
index 60044fb863..c3378c3934 100644
--- a/astrbot/core/provider/manager.py
+++ b/astrbot/core/provider/manager.py
@@ -455,6 +455,14 @@ def dynamic_import_provider(self, type: str) -> None:
                 from .sources.mirarouter_source import (
                     ProviderMiraRouter as ProviderMiraRouter,
                 )
+            case (
+                "opencode_go_chat_completion"
+                | "opencode_go_responses"
+                | "opencode_go_messages"
+            ):
+                from .sources.opencode_go_source import (
+                    ProviderOpenCodeGo as ProviderOpenCodeGo,
+                )
             case "openrouter_chat_completion":
                 from .sources.openrouter_source import (
                     ProviderOpenRouter as ProviderOpenRouter,
diff --git a/astrbot/core/provider/sources/anthropic_source.py b/astrbot/core/provider/sources/anthropic_source.py
index 5a21737b44..fbc861646d 100644
--- a/astrbot/core/provider/sources/anthropic_source.py
+++ b/astrbot/core/provider/sources/anthropic_source.py
@@ -812,6 +812,8 @@ async def text_chat(
         model = model or self.get_model()
 
         payloads = {"messages": new_messages, "model": model}
+        if extra_headers := kwargs.get("extra_headers"):
+            payloads["extra_headers"] = extra_headers
         if func_tool and not func_tool.empty():
             payloads["tool_choice"] = tool_choice
 
@@ -884,6 +886,8 @@ async def text_chat_stream(
         model = model or self.get_model()
 
         payloads = {"messages": new_messages, "model": model}
+        if extra_headers := kwargs.get("extra_headers"):
+            payloads["extra_headers"] = extra_headers
         if func_tool and not func_tool.empty():
             payloads["tool_choice"] = tool_choice
 
diff --git a/astrbot/core/provider/sources/openai_responses_source.py b/astrbot/core/provider/sources/openai_responses_source.py
index c5cb9bdb82..aa8cf7a1aa 100644
--- a/astrbot/core/provider/sources/openai_responses_source.py
+++ b/astrbot/core/provider/sources/openai_responses_source.py
@@ -241,6 +241,7 @@ async def _prepare_chat_payload(
         tool_calls_result: ToolCallsResult | list[ToolCallsResult] | None = None,
         model: str | None = None,
         extra_user_content_parts: list[ContentPart] | None = None,
+        extra_headers: dict[str, str] | None = None,
         **kwargs: Any,
     ) -> tuple[dict, list[dict]]:
         """Build a stateless Responses API payload and replayable context.
@@ -254,6 +255,7 @@ async def _prepare_chat_payload(
             tool_calls_result: Function calls and their returned outputs.
             model: Optional per-request model override.
             extra_user_content_parts: Additional user content blocks.
+            extra_headers: HTTP headers applied only to this request.
             **kwargs: Reserved provider request arguments.
 
         Returns:
@@ -291,6 +293,8 @@ async def _prepare_chat_payload(
         }
         if system_prompt:
             payloads["instructions"] = system_prompt
+        if extra_headers:
+            payloads["extra_headers"] = extra_headers
 
         return payloads, context_query
 
diff --git a/astrbot/core/provider/sources/openai_source.py b/astrbot/core/provider/sources/openai_source.py
index 15fd6b72f4..9a05600015 100644
--- a/astrbot/core/provider/sources/openai_source.py
+++ b/astrbot/core/provider/sources/openai_source.py
@@ -950,6 +950,7 @@ async def _prepare_chat_payload(
         tool_calls_result: ToolCallsResult | list[ToolCallsResult] | None = None,
         model: str | None = None,
         extra_user_content_parts: list[ContentPart] | None = None,
+        extra_headers: dict[str, str] | None = None,
         **kwargs,
     ) -> tuple:
         """准备聊天所需的有效载荷和上下文"""
@@ -987,6 +988,8 @@ async def _prepare_chat_payload(
         model = model or self.get_model()
 
         payloads = {"messages": context_query, "model": model}
+        if extra_headers:
+            payloads["extra_headers"] = extra_headers
 
         self._finally_convert_payload(payloads)
 
diff --git a/astrbot/core/provider/sources/opencode_go_source.py b/astrbot/core/provider/sources/opencode_go_source.py
new file mode 100644
index 0000000000..ebeb6b2b56
--- /dev/null
+++ b/astrbot/core/provider/sources/opencode_go_source.py
@@ -0,0 +1,194 @@
+import hashlib
+from collections.abc import AsyncGenerator
+from uuid import uuid4
+
+from astrbot.core.provider.entities import LLMResponse
+from astrbot.core.provider.provider import Provider
+
+from ..register import register_provider_adapter
+from .anthropic_source import ProviderAnthropic
+from .openai_responses_source import ProviderOpenAIResponses
+from .openai_source import ProviderOpenAIOfficial
+
+OPENCODE_GO_API_BASE = "https://opencode.ai/zen/go/v1"
+
+
+@register_provider_adapter(
+    "opencode_go_chat_completion", "OpenCode Go Chat Completions Provider Adapter"
+)
+class ProviderOpenCodeGo(Provider):
+    """Send Go requests using the explicitly selected protocol adapter."""
+
+    ADAPTER: type[Provider] = ProviderOpenAIOfficial
+
+    def __init__(self, provider_config: dict, provider_settings: dict) -> None:
+        super().__init__(provider_config, provider_settings)
+        self.set_model(provider_config.get("model") or "unknown")
+        config = dict(provider_config)
+        config["api_base"] = config.get("api_base") or OPENCODE_GO_API_BASE
+        config["model"] = self.get_model().removeprefix("opencode-go/")
+        config["custom_headers"] = {
+            key: value
+            for key, value in self.request_headers.items()
+            if key.lower() != "x-opencode-session"
+        }
+        self.delegate = self.ADAPTER(config, provider_settings)
+
+    def get_current_key(self) -> str:
+        return self.delegate.get_current_key()
+
+    def get_keys(self) -> list[str]:
+        return self.delegate.get_keys()
+
+    def set_key(self, key: str) -> None:
+        self.delegate.set_key(key)
+
+    async def get_models(self) -> list[str]:
+        return await self.delegate.get_models()
+
+    async def text_chat(
+        self,
+        prompt=None,
+        session_id=None,
+        image_urls=None,
+        audio_urls=None,
+        func_tool=None,
+        contexts=None,
+        system_prompt=None,
+        tool_calls_result=None,
+        model=None,
+        extra_user_content_parts=None,
+        tool_choice="auto",
+        **kwargs,
+    ) -> LLMResponse:
+        """Send a request using the conversation UUID as the session identity.
+
+        Args:
+            prompt: Current user prompt.
+            session_id: Deprecated provider argument, forwarded for compatibility.
+            image_urls: Images attached to the request.
+            audio_urls: Audio attached to the request.
+            func_tool: Available function tools.
+            contexts: Conversation history.
+            system_prompt: System instructions.
+            tool_calls_result: Results of previous tool calls.
+            model: Optional per-request model override.
+            extra_user_content_parts: Additional user content blocks.
+            tool_choice: Whether tool use is automatic or required.
+            **kwargs: Optional conversation_id (AstrBot conversation UUID) and
+                additional arguments forwarded to the protocol adapter.
+
+        Returns:
+            The normalized model response.
+        """
+        model = (model or self.get_model()).removeprefix("opencode-go/")
+        # Normalize UA casing for SDK merging; blank overrides keep client defaults.
+        extra_headers = {
+            ("User-Agent" if key.lower() == "user-agent" else key): value
+            for key, value in (kwargs.pop("extra_headers", None) or {}).items()
+            if key.lower() != "x-opencode-session"
+            and (key.lower() != "user-agent" or str(value).strip())
+        }
+        # Calls without a conversation (such as connection tests) are independent.
+        extra_headers["x-opencode-session"] = hashlib.sha256(
+            (kwargs.pop("conversation_id", None) or uuid4().hex).encode()
+        ).hexdigest()
+        return await self.delegate.text_chat(
+            prompt=prompt,
+            session_id=session_id,
+            image_urls=image_urls,
+            audio_urls=audio_urls,
+            func_tool=func_tool,
+            contexts=contexts,
+            system_prompt=system_prompt,
+            tool_calls_result=tool_calls_result,
+            model=model,
+            extra_user_content_parts=extra_user_content_parts,
+            tool_choice=tool_choice,
+            extra_headers=extra_headers,
+            **kwargs,
+        )
+
+    async def text_chat_stream(
+        self,
+        prompt=None,
+        session_id=None,
+        image_urls=None,
+        audio_urls=None,
+        func_tool=None,
+        contexts=None,
+        system_prompt=None,
+        tool_calls_result=None,
+        model=None,
+        extra_user_content_parts=None,
+        tool_choice="auto",
+        **kwargs,
+    ) -> AsyncGenerator[LLMResponse, None]:
+        """Stream a response with the same session identity as non-streaming calls.
+
+        Args:
+            prompt: Current user prompt.
+            session_id: Deprecated provider argument, forwarded for compatibility.
+            image_urls: Images attached to the request.
+            audio_urls: Audio attached to the request.
+            func_tool: Available function tools.
+            contexts: Conversation history.
+            system_prompt: System instructions.
+            tool_calls_result: Results of previous tool calls.
+            model: Optional per-request model override.
+            extra_user_content_parts: Additional user content blocks.
+            tool_choice: Whether tool use is automatic or required.
+            **kwargs: Optional conversation_id (AstrBot conversation UUID) and
+                additional arguments forwarded to the protocol adapter.
+
+        Yields:
+            Normalized response chunks.
+        """
+        model = (model or self.get_model()).removeprefix("opencode-go/")
+        # Normalize UA casing for SDK merging; blank overrides keep client defaults.
+        extra_headers = {
+            ("User-Agent" if key.lower() == "user-agent" else key): value
+            for key, value in (kwargs.pop("extra_headers", None) or {}).items()
+            if key.lower() != "x-opencode-session"
+            and (key.lower() != "user-agent" or str(value).strip())
+        }
+        extra_headers["x-opencode-session"] = hashlib.sha256(
+            (kwargs.pop("conversation_id", None) or uuid4().hex).encode()
+        ).hexdigest()
+        async for response in self.delegate.text_chat_stream(
+            prompt=prompt,
+            session_id=session_id,
+            image_urls=image_urls,
+            audio_urls=audio_urls,
+            func_tool=func_tool,
+            contexts=contexts,
+            system_prompt=system_prompt,
+            tool_calls_result=tool_calls_result,
+            model=model,
+            extra_user_content_parts=extra_user_content_parts,
+            tool_choice=tool_choice,
+            extra_headers=extra_headers,
+            **kwargs,
+        ):
+            yield response
+
+    async def terminate(self) -> None:
+        await self.delegate.terminate()
+
+
+@register_provider_adapter(
+    "opencode_go_responses", "OpenCode Go Responses Provider Adapter"
+)
+class ProviderOpenCodeGoResponses(ProviderOpenCodeGo):
+    """Use Go's Responses endpoint for user-selected models."""
+
+    ADAPTER = ProviderOpenAIResponses
+
+
+@register_provider_adapter(
+    "opencode_go_messages", "OpenCode Go Messages Provider Adapter"
+)
+class ProviderOpenCodeGoMessages(ProviderOpenCodeGo):
+    """Use Go's Messages endpoint for user-selected models."""
+
+    ADAPTER = ProviderAnthropic
diff --git a/astrbot/core/star/context.py b/astrbot/core/star/context.py
index b4f6e61c48..becde7100c 100644
--- a/astrbot/core/star/context.py
+++ b/astrbot/core/star/context.py
@@ -245,6 +245,7 @@ async def tool_loop_agent(
                 stream: bool - whether to stream the LLM response
                 agent_hooks: BaseAgentRunHooks[AstrAgentContext] - hooks to run during agent execution
                 agent_context: AstrAgentContext - context to use for the agent
+                conversation_id: str - stable identity for a plugin-managed conversation; without it each call gets a random one
 
                 other kwargs will be DIRECTLY passed to the runner.reset() method
 
diff --git a/docs/en/providers/opencode-go.md b/docs/en/providers/opencode-go.md
new file mode 100644
index 0000000000..a70b1e4c40
--- /dev/null
+++ b/docs/en/providers/opencode-go.md
@@ -0,0 +1,30 @@
+# Connect OpenCode Go
+
+[OpenCode Go](https://opencode.ai/docs/go/) is a model subscription service for coding agents.
+
+## Get an API Key
+
+Open the [OpenCode console](https://opencode.ai/auth), subscribe to Go, and copy your API key.
+
+## Configure AstrBot
+
+Open the AstrBot dashboard and go to **Providers → Add Provider**. Select **OpenCode Go Chat Completions**, **OpenCode Go Responses**, or **OpenCode Go Messages** according to the model's API format in the [OpenCode Go documentation](https://opencode.ai/docs/go/#endpoints).
+
+| Field | Value |
+| --- | --- |
+| API Base URL | `https://opencode.ai/zen/go/v1` |
+| API Key | The API key obtained from the OpenCode console |
+
+Save the provider, then open its card and add the models you want to use.
+
+## Request Headers
+
+The default `User-Agent` is `astrbot/<version>`. Non-blank values in per-request `extra_headers` take priority over provider `custom_headers`, which take priority over the default. User-Agent header names are case-insensitive; blank values fall back to the next level.
+
+`x-opencode-session` is generated automatically from the conversation ID and cannot be overridden. Direct calls without a conversation ID get an independent session.
+
+## Set as Default
+
+Go to **Settings → Provider Settings**, select the OpenCode Go model you just added as the default chat model, and save the configuration.
+
+For supported models, usage requirements, and limits, see the [OpenCode Go documentation](https://opencode.ai/docs/go/).
diff --git a/docs/zh/providers/opencode-go.md b/docs/zh/providers/opencode-go.md
new file mode 100644
index 0000000000..5aab49d745
--- /dev/null
+++ b/docs/zh/providers/opencode-go.md
@@ -0,0 +1,30 @@
+# 接入 OpenCode Go
+
+[OpenCode Go](https://opencode.ai/docs/go/) 是面向编程代理的模型订阅服务。
+
+## 获取 API Key
+
+前往 [OpenCode 控制台](https://opencode.ai/auth)，订阅 Go 并复制 API Key。
+
+## 在 AstrBot 中配置
+
+打开 AstrBot 管理面板，进入 **服务提供商 → 新增提供商**，根据 [OpenCode Go 文档](https://opencode.ai/docs/go/#endpoints)中模型的接口类型选择 **OpenCode Go Chat Completions**、**OpenCode Go Responses** 或 **OpenCode Go Messages**。
+
+| 配置项 | 值 |
+| --- | --- |
+| API Base URL | `https://opencode.ai/zen/go/v1` |
+| API Key | 在 OpenCode 控制台获取的 API Key |
+
+保存后，点击提供商卡片，添加需要使用的模型。
+
+## 请求头
+
+默认 `User-Agent` 为 `astrbot/<版本>`。非空白值的优先级为：单次请求的 `extra_headers` → 提供商的 `custom_headers` → 默认值。User-Agent 键名不区分大小写，空白值回退到下一层。
+
+`x-opencode-session` 根据对话 ID 自动生成，不允许手动覆盖。未传入对话 ID 的直接调用使用独立会话标识。
+
+## 设为默认模型
+
+进入 **配置文件 → 提供商设置**，将「默认聊天模型」设置为刚刚添加的 OpenCode Go 模型，然后保存配置。
+
+支持的模型、使用要求和额度请参阅 [OpenCode Go 文档](https://opencode.ai/docs/go/)。
diff --git a/tests/test_opencode_go_source.py b/tests/test_opencode_go_source.py
new file mode 100644
index 0000000000..3b494f4e57
--- /dev/null
+++ b/tests/test_opencode_go_source.py
@@ -0,0 +1,410 @@
+import asyncio
+import copy
+import hashlib
+import json
+from functools import partial
+from uuid import uuid4
+
+import httpx
+import pytest
+from anthropic import _base_client as anthropic_base_client
+from openai import _base_client as openai_base_client
+
+from astrbot.core.agent.context.config import ContextConfig
+from astrbot.core.agent.context.manager import ContextManager
+from astrbot.core.agent.message import Message
+from astrbot.core.config.default import CONFIG_METADATA_2
+from astrbot.core.provider.headers import DEFAULT_USER_AGENT
+from astrbot.core.provider.sources.anthropic_source import ProviderAnthropic
+from astrbot.core.provider.sources.openai_source import ProviderOpenAIOfficial
+from astrbot.core.provider.sources.opencode_go_source import (
+    ProviderOpenCodeGo,
+    ProviderOpenCodeGoMessages,
+    ProviderOpenCodeGoResponses,
+)
+
+GO_PROTOCOL_CASES = [
+    (ProviderOpenCodeGo, "chat/completions"),
+    (ProviderOpenCodeGoResponses, "responses"),
+    (ProviderOpenCodeGoMessages, "messages"),
+]
+
+
+@pytest.fixture
+def go_http(monkeypatch):
+    """Capture real SDK HTTP requests with deterministic protocol responses."""
+    requests = []
+
+    async def handle(request, *, httpx_module):
+        requests.append(request)
+        await asyncio.sleep(0)
+        if request.method == "GET":
+            return httpx_module.Response(200, json={"data": [{"id": "kimi-k2.6"}]})
+        body = json.loads(request.content)
+        model = body["model"]
+        if request.url.path.endswith("/chat/completions"):
+            response = {
+                "id": "chat-1",
+                "object": "chat.completion",
+                "created": 1,
+                "model": model,
+                "choices": [
+                    {
+                        "index": 0,
+                        "message": {"role": "assistant", "content": "ok"},
+                        "finish_reason": "stop",
+                    }
+                ],
+            }
+            events = [
+                {
+                    **response,
+                    "object": "chat.completion.chunk",
+                    "choices": [
+                        {
+                            "index": 0,
+                            "delta": {"role": "assistant", "content": "ok"},
+                            "finish_reason": "stop",
+                        }
+                    ],
+                }
+            ]
+        elif request.url.path.endswith("/responses"):
+            response = {
+                "id": "resp-1",
+                "object": "response",
+                "created_at": 1,
+                "model": model,
+                "status": "completed",
+                "parallel_tool_calls": True,
+                "tool_choice": "auto",
+                "tools": [],
+                "output": [
+                    {
+                        "type": "message",
+                        "id": "msg-1",
+                        "role": "assistant",
+                        "status": "completed",
+                        "content": [
+                            {"type": "output_text", "text": "ok", "annotations": []}
+                        ],
+                    }
+                ],
+            }
+            events = [
+                {
+                    "type": "response.completed",
+                    "response": response,
+                    "sequence_number": 0,
+                }
+            ]
+        else:
+            assert request.url.path == "/zen/go/v1/messages"
+            response = {
+                "id": "msg-1",
+                "type": "message",
+                "role": "assistant",
+                "model": model,
+                "content": [{"type": "text", "text": "ok"}],
+                "stop_reason": "end_turn",
+                "stop_sequence": None,
+                "usage": {"input_tokens": 1, "output_tokens": 1},
+            }
+            events = [
+                {
+                    "type": "message_start",
+                    "message": {**response, "content": [], "stop_reason": None},
+                },
+                {
+                    "type": "content_block_start",
+                    "index": 0,
+                    "content_block": {"type": "text", "text": ""},
+                },
+                {
+                    "type": "content_block_delta",
+                    "index": 0,
+                    "delta": {"type": "text_delta", "text": "ok"},
+                },
+                {"type": "content_block_stop", "index": 0},
+                {
+                    "type": "message_delta",
+                    "delta": {"stop_reason": "end_turn", "stop_sequence": None},
+                    "usage": {"output_tokens": 1},
+                },
+                {"type": "message_stop"},
+            ]
+        if body.get("stream"):
+            content = "".join(
+                f"event: {event.get('type', 'message')}\ndata: {json.dumps(event)}\n\n"
+                for event in events
+            )
+            return httpx_module.Response(
+                200, text=content, headers={"Content-Type": "text/event-stream"}
+            )
+        return httpx_module.Response(200, json=response)
+
+    def client(provider, _config):
+        sdk = (
+            anthropic_base_client
+            if isinstance(provider, ProviderAnthropic)
+            else openai_base_client
+        )
+        httpx_module = getattr(sdk, "httpx", getattr(sdk, "httpx2", httpx))
+        return httpx_module.AsyncClient(
+            transport=httpx_module.MockTransport(
+                partial(handle, httpx_module=httpx_module)
+            )
+        )
+
+    monkeypatch.setattr(ProviderOpenAIOfficial, "_create_http_client", client)
+    monkeypatch.setattr(ProviderAnthropic, "_create_http_client", client)
+    return requests
+
+
+@pytest.mark.asyncio
+@pytest.mark.parametrize("streaming", [False, True])
+@pytest.mark.parametrize("provider_class,endpoint", GO_PROTOCOL_CASES)
+async def test_go_http_identity_and_concurrent_sessions(
+    go_http, provider_class, endpoint, streaming
+):
+    model = "new-model-without-routing-metadata"
+    provider = provider_class(
+        {
+            "key": ["test-key"],
+            "custom_headers": {
+                "user-agent": "configured/1.0",
+                "X-OpenCode-Session": "wrong",
+                "X-Custom": "keep",
+            },
+        },
+        {},
+    )
+    sessions = [str(uuid4()) for _ in range(4)]
+    expected_session_ids = [
+        hashlib.sha256(conversation_id.encode()).hexdigest()
+        for conversation_id in sessions
+    ]
+    expected_user_agents = {
+        session_id: f"request/{conversation_id}"
+        for session_id, conversation_id in zip(expected_session_ids, sessions)
+    }
+
+    async def send(conversation_id):
+        kwargs = {
+            "prompt": "Write a Python function",
+            "conversation_id": conversation_id,
+            "model": f"opencode-go/{model}",
+            "extra_headers": {
+                "X-Request-Test": "request-header",
+                "uSeR-aGeNt": f"request/{conversation_id}",
+                "X-OPENCODE-SESSION": "wrong-request-session",
+            },
+        }
+        if streaming:
+            result = [item async for item in provider.text_chat_stream(**kwargs)]
+            assert any(item.completion_text == "ok" for item in result)
+        else:
+            assert (await provider.text_chat(**kwargs)).completion_text == "ok"
+
+    try:
+        await asyncio.gather(*(send(conversation_id) for conversation_id in sessions))
+        await send(sessions[0])
+        assert len(go_http) == 5
+        actual_session_ids = [r.headers["x-opencode-session"] for r in go_http]
+        assert set(actual_session_ids) == set(expected_session_ids)
+        assert actual_session_ids[-1] == expected_session_ids[0]
+        assert actual_session_ids.count(expected_session_ids[0]) == 2
+        for request in go_http:
+            assert request.url.path == f"/zen/go/v1/{endpoint}"
+            assert request.headers.get_list("user-agent") == [
+                expected_user_agents[request.headers["x-opencode-session"]]
+            ]
+            assert len(request.headers.get_list("x-opencode-session")) == 1
+            assert request.headers["x-custom"] == "keep"
+            assert request.headers["x-request-test"] == "request-header"
+            body = json.loads(request.content)
+            assert body["model"] == model
+            assert "extra_headers" not in body
+            assert "x-opencode-session" not in body
+            assert "X-Request-Test" not in body
+    finally:
+        await provider.terminate()
+
+
+@pytest.mark.asyncio
+@pytest.mark.parametrize("streaming", [False, True])
+@pytest.mark.parametrize("provider_class,endpoint", GO_PROTOCOL_CASES)
+@pytest.mark.parametrize(
+    "custom_headers,extra_headers,expected_user_agent",
+    [
+        (None, None, DEFAULT_USER_AGENT),
+        ({}, {}, DEFAULT_USER_AGENT),
+        ({"user-agent": "configured/1.0"}, {}, "configured/1.0"),
+        ({"USER-AGENT": " "}, {}, DEFAULT_USER_AGENT),
+        ({}, {"user-agent": "request/1.0"}, "request/1.0"),
+        (
+            {"uSeR-aGeNt": "configured/1.0"},
+            {"User-Agent": "request/1.0"},
+            "request/1.0",
+        ),
+        (
+            {"User-Agent": "configured/1.0"},
+            {"USER-AGENT": "request/1.0"},
+            "request/1.0",
+        ),
+        (
+            {"User-Agent": "configured/1.0"},
+            {"User-Agent": " "},
+            "configured/1.0",
+        ),
+        (
+            {"User-Agent": "configured/1.0"},
+            {"uSeR-aGeNt": ""},
+            "configured/1.0",
+        ),
+        ({"user-agent": " "}, {"USER-AGENT": " "}, DEFAULT_USER_AGENT),
+        (
+            {"user-agent": "configured/1.0", "X-OpenCode-Session": "wrong"},
+            {"X-Request-Test": "keep"},
+            "configured/1.0",
+        ),
+        (
+            {"User-Agent": "first/1.0", "USER-AGENT": "configured/1.0"},
+            {
+                "user-agent": "first/1.0",
+                "User-Agent": "request/1.0",
+                "USER-AGENT": " ",
+            },
+            "request/1.0",
+        ),
+    ],
+)
+async def test_go_user_agent_defaults_and_overrides(
+    go_http,
+    provider_class,
+    endpoint,
+    streaming,
+    custom_headers,
+    extra_headers,
+    expected_user_agent,
+):
+    """Keep one UA per request without changing configuration or client defaults."""
+    config = {"key": ["test-key"], "custom_headers": custom_headers}
+    original_config = copy.deepcopy(config)
+    original_extra_headers = copy.deepcopy(extra_headers)
+    provider = provider_class(config, {})
+    default_headers = dict(provider.delegate.request_headers)
+    try:
+        kwargs = {
+            "prompt": "Write code",
+            "conversation_id": "test-conversation",
+            "extra_headers": extra_headers,
+        }
+        if streaming:
+            result = [item async for item in provider.text_chat_stream(**kwargs)]
+            assert any(item.completion_text == "ok" for item in result)
+        else:
+            assert (await provider.text_chat(**kwargs)).completion_text == "ok"
+        assert len(go_http) == 1
+        request = go_http[0]
+        assert request.url.path == f"/zen/go/v1/{endpoint}"
+        assert request.headers.get_list("user-agent") == [expected_user_agent]
+        assert request.headers.get_list("x-opencode-session") == [
+            hashlib.sha256(b"test-conversation").hexdigest()
+        ]
+        assert "extra_headers" not in json.loads(request.content)
+        await provider.get_models()
+        assert go_http[-1].headers.get_list("user-agent") == [
+            provider.request_headers["User-Agent"]
+        ]
+        assert "x-opencode-session" not in go_http[-1].headers
+        assert provider.delegate.request_headers == default_headers
+        assert config == original_config
+        assert extra_headers == original_extra_headers
+    finally:
+        await provider.terminate()
+
+
+@pytest.mark.asyncio
+@pytest.mark.parametrize("provider_class,endpoint", GO_PROTOCOL_CASES)
+async def test_go_model_changes_preserve_selected_protocol(
+    go_http, provider_class, endpoint
+):
+    provider = provider_class({"key": ["test-key"]}, {})
+    try:
+        assert await provider.get_models() == ["kimi-k2.6"]
+        for model in ["kimi-k2.6", "gpt-5.6-luna", "minimax-m3"]:
+            provider.set_model(model)
+            await provider.text_chat(prompt="Write code")
+        assert [r.url.path for r in go_http[1:]] == [f"/zen/go/v1/{endpoint}"] * 3
+        assert [json.loads(r.content)["model"] for r in go_http[1:]] == [
+            "kimi-k2.6",
+            "gpt-5.6-luna",
+            "minimax-m3",
+        ]
+        assert len({r.headers["x-opencode-session"] for r in go_http[1:]}) == 3
+    finally:
+        await provider.terminate()
+
+
+@pytest.mark.asyncio
+async def test_go_summary_preserves_conversation_conversation_id(go_http):
+    provider = ProviderOpenCodeGo({"key": ["test-key"]}, {})
+    conversation_id = str(uuid4())
+    manager = ContextManager(
+        ContextConfig(llm_compress_provider=provider, llm_compress_keep_recent_ratio=0),
+        conversation_id=conversation_id,
+    )
+    try:
+        await provider.text_chat(prompt="Write code", conversation_id=conversation_id)
+        await manager.compressor(
+            [
+                Message(role="user", content="Write code"),
+                Message(role="assistant", content="Here is the code"),
+            ]
+        )
+        assert len(go_http) == 2
+        assert (
+            go_http[0].headers["x-opencode-session"]
+            == go_http[1].headers["x-opencode-session"]
+        )
+    finally:
+        await provider.terminate()
+
+
+@pytest.mark.parametrize(
+    "name,provider_type,provider_class",
+    [
+        ("Chat Completions", "opencode_go_chat_completion", ProviderOpenCodeGo),
+        ("Responses", "opencode_go_responses", ProviderOpenCodeGoResponses),
+        ("Messages", "opencode_go_messages", ProviderOpenCodeGoMessages),
+    ],
+)
+def test_go_templates(name, provider_type, provider_class):
+    from astrbot.core.provider.register import provider_cls_map
+
+    template = CONFIG_METADATA_2["provider_group"]["metadata"]["provider"][
+        "config_template"
+    ][f"OpenCode Go {name}"]
+    assert template["type"] == provider_type
+    assert provider_cls_map[provider_type].cls_type is provider_class
+    assert template["api_base"] == "https://opencode.ai/zen/go/v1"
+
+
+@pytest.mark.asyncio
+async def test_go_conversation_switch_within_same_umo(go_http):
+    provider = ProviderOpenCodeGo({"key": ["test-key"]}, {})
+    first_cid, second_cid = str(uuid4()), str(uuid4())
+    try:
+        for cid in [first_cid, second_cid, first_cid]:
+            await provider.text_chat(
+                prompt="Write code",
+                session_id="qq:GroupMessage:456",
+                conversation_id=cid,
+            )
+        identities = [request.headers["x-opencode-session"] for request in go_http]
+        assert identities[0] == identities[2]
+        assert identities[0] != identities[1]
+        assert identities[0] == hashlib.sha256(first_cid.encode()).hexdigest()
+    finally:
+        await provider.terminate()
diff --git a/tests/test_tool_loop_agent_runner.py b/tests/test_tool_loop_agent_runner.py
index 180e0edf2d..1e84ff4942 100644
--- a/tests/test_tool_loop_agent_runner.py
+++ b/tests/test_tool_loop_agent_runner.py
@@ -5,6 +5,7 @@
 from types import SimpleNamespace
 from typing import Any, cast
 from unittest.mock import AsyncMock, MagicMock
+from uuid import uuid4
 
 import pytest
 
@@ -20,10 +21,13 @@
 from astrbot.core.agent.tool import FunctionTool, ToolSet
 from astrbot.core.astr_agent_run_util import run_agent
 from astrbot.core.astr_agent_tool_exec import FunctionToolExecutor
+from astrbot.core.db.po import Conversation
 from astrbot.core.exceptions import EmptyModelOutputError
 from astrbot.core.message.message_event_result import MessageChain
+from astrbot.core.platform.astr_message_event import AstrMessageEvent
 from astrbot.core.provider.entities import LLMResponse, ProviderRequest, TokenUsage
 from astrbot.core.provider.provider import Provider
+from astrbot.core.star.context import Context
 
 
 class MockProvider(Provider):
@@ -619,6 +623,130 @@ async def snapshot_context_manager(messages, trusted_token_usage=0):
     assert "工具执行结果" in tool_messages[0].content
 
 
+@pytest.mark.asyncio
+@pytest.mark.parametrize("streaming", [False, True])
+@pytest.mark.parametrize("persistent", [False, True])
+@pytest.mark.parametrize("tool_schema_mode", ["full", "skills_like"])
+async def test_conversation_identity_is_stable_within_each_run(
+    runner,
+    mock_provider,
+    tool_set,
+    mock_tool_executor,
+    mock_hooks,
+    streaming,
+    persistent,
+    tool_schema_mode,
+):
+    """Keep one identity through tools, requery repair, compression, and reset."""
+    conversation = (
+        Conversation(platform_id="test", user_id="user", cid=str(uuid4()))
+        if persistent
+        else None
+    )
+    identities = []
+    for _ in range(2):
+        tool_response = LLMResponse(
+            role="assistant",
+            tools_call_name=["test_tool"],
+            tools_call_args=[{"query": "test"}],
+            tools_call_ids=["call_identity"],
+        )
+        responses = [tool_response]
+        if tool_schema_mode == "skills_like":
+            responses.extend([LLMResponse(role="assistant"), tool_response])
+        responses.extend(
+            [
+                LLMResponse(role="assistant", completion_text="final"),
+                LLMResponse(role="assistant", completion_text="summary"),
+            ]
+        )
+        mock_provider.text_chat = AsyncMock(side_effect=responses)
+        request = ProviderRequest(
+            prompt="Run the tool",
+            func_tool=tool_set,
+            conversation=conversation,
+        )
+        await runner.reset(
+            provider=mock_provider,
+            request=request,
+            run_context=ContextWrapper(context=None),
+            tool_executor=mock_tool_executor,
+            agent_hooks=mock_hooks,
+            streaming=streaming,
+            tool_schema_mode=tool_schema_mode,
+            llm_compress_provider=mock_provider,
+            llm_compress_keep_recent_ratio=0,
+        )
+        async for _ in runner.step_until_done(3):
+            pass
+        assert runner.done()
+        assert any(message.role == "tool" for message in runner.run_context.messages)
+        await runner.request_context_manager.compressor(runner.run_context.messages)
+
+        calls = mock_provider.text_chat.call_args_list
+        assert len(calls) == (5 if tool_schema_mode == "skills_like" else 3)
+        identity = calls[0].kwargs["conversation_id"]
+        assert identity
+        assert all(call.kwargs["conversation_id"] == identity for call in calls)
+        assert request.conversation is conversation
+        if conversation:
+            assert identity == conversation.cid
+        identities.append(identity)
+
+    assert (identities[0] == identities[1]) is persistent
+
+
+@pytest.mark.asyncio
+async def test_tool_loop_agent_passes_explicit_conversation_id():
+    """Context.tool_loop_agent forwards conversation_id through runner to provider."""
+    provider = MockProvider()
+    context = Context(
+        event_queue=AsyncMock(),
+        config=MagicMock(),
+        db=MagicMock(),
+        provider_manager=SimpleNamespace(
+            get_provider_by_id=AsyncMock(return_value=provider)
+        ),
+        platform_manager=MagicMock(),
+        conversation_manager=MagicMock(),
+        message_history_manager=MagicMock(),
+        persona_manager=MagicMock(),
+        astrbot_config_mgr=MagicMock(),
+        knowledge_base_manager=MagicMock(),
+        cron_manager=MagicMock(),
+    )
+    event = MagicMock(spec=AstrMessageEvent)
+    event.unified_msg_origin = "test_umo"
+    seen: list[str | None] = []
+
+    async def text_chat(**kwargs):
+        seen.append(kwargs.get("conversation_id"))
+        return LLMResponse(role="assistant", completion_text="done")
+
+    provider.text_chat = text_chat
+
+    async def run_once(conversation_id: str | None = None) -> None:
+        kwargs = {"conversation_id": conversation_id} if conversation_id else {}
+        resp = await context.tool_loop_agent(
+            event=event,
+            chat_provider_id="provider-id",
+            prompt="hi",
+            **kwargs,
+        )
+        assert resp.completion_text == "done"
+
+    await run_once("stable-id")
+    await run_once("stable-id")
+    assert seen == ["stable-id", "stable-id"]
+    await run_once("other-id")
+    assert seen[-1] == "other-id"
+    transient = len(seen)
+    await run_once()
+    await run_once()
+    assert seen[transient] and seen[transient + 1]
+    assert seen[transient] != seen[transient + 1]
+
+
 @pytest.mark.asyncio
 async def test_normal_completion_without_max_step(
     runner, mock_provider, provider_request, mock_tool_executor, mock_hooks