universal-mcp-agents 0.1.2__tar.gz → 0.1.4__tar.gz

{universal_mcp_agents-0.1.2 → universal_mcp_agents-0.1.4}/PKG-INFO

@@ -1,18 +1,18 @@
 Metadata-Version: 2.4
 Name: universal-mcp-agents
-Version: 0.1.2
+Version: 0.1.4
 Summary: Add your description here
 Project-URL: Homepage, https://github.com/universal-mcp/applications
 Project-URL: Repository, https://github.com/universal-mcp/applications
 Author-email: Manoj Bajaj <manojbajaj95@gmail.com>
 License: MIT
-Requires-Python: >=3.13
+Requires-Python: >=3.11
 Requires-Dist: langchain-anthropic>=0.3.19
 Requires-Dist: langchain-google-genai>=2.1.10
 Requires-Dist: langchain-openai>=0.3.32
 Requires-Dist: langgraph>=0.6.6
-Requires-Dist: universal-mcp
-Requires-Dist: universal-mcp-applications>=0.1.2
+Requires-Dist: universal-mcp-applications>=0.1.4
+Requires-Dist: universal-mcp>=0.1.24rc17
 Provides-Extra: dev
 Requires-Dist: pre-commit; extra == 'dev'
 Requires-Dist: ruff; extra == 'dev'

{universal_mcp_agents-0.1.2 → universal_mcp_agents-0.1.4}/pyproject.toml

@@ -6,20 +6,20 @@ build-backend = "hatchling.build"
 
 [project]
 name = "universal-mcp-agents"
-version = "0.1.2"
+version = "0.1.4"
 description = "Add your description here"
 readme = "README.md"
 authors = [
     { name = "Manoj Bajaj", email = "manojbajaj95@gmail.com" }
 ]
-requires-python = ">=3.13"
+requires-python = ">=3.11"
 dependencies = [
     "langchain-anthropic>=0.3.19",
     "langchain-google-genai>=2.1.10",
     "langchain-openai>=0.3.32",
     "langgraph>=0.6.6",
-    "universal-mcp",
-    "universal-mcp-applications>=0.1.2",
+    "universal-mcp >= 0.1.24rc17",
+    "universal-mcp-applications>=0.1.4",
 ]
 
 [project.license]

{universal_mcp_agents-0.1.2 → universal_mcp_agents-0.1.4}/src/evals/evaluators.py

@@ -11,13 +11,17 @@ def exact_match_evaluator(run: Run, example: Example | None = None) -> Evaluatio
     and the expected output from the dataset.
     """
     if example is None or "expected_output" not in example.outputs:
-        return EvaluationResult(key="exact_match", score=0, comment="No expected output provided.")
+        return EvaluationResult(
+            key="exact_match", score=0, comment="No expected output provided."
+        )
 
     # The agent's response might be in a list of messages
     agent_response_raw = run.outputs.get("output", "")
     if isinstance(agent_response_raw, list):
         # Extract text from the last dictionary in the list
-        agent_response = agent_response_raw[-1].get("text", "").strip() if agent_response_raw else ""
+        agent_response = (
+            agent_response_raw[-1].get("text", "").strip() if agent_response_raw else ""
+        )
     else:
         agent_response = str(agent_response_raw).strip()
 

{universal_mcp_agents-0.1.2 → universal_mcp_agents-0.1.4}/src/evals/run.py

@@ -34,7 +34,9 @@ def get_agent(agent_name: str) -> BaseAgent:
     elif agent_name == "auto":
         return AutoAgent(name="auto-agent", **common_params)
     else:
-        raise ValueError(f"Unknown agent: {agent_name}. Available agents: simple, react, auto")
+        raise ValueError(
+            f"Unknown agent: {agent_name}. Available agents: simple, react, auto"
+        )
 
 
 # 2. Evaluator Registry
@@ -50,7 +52,9 @@ def get_evaluator(evaluator_name: str) -> RunEvaluator:
     """
     evaluator = EVALUATORS.get(evaluator_name)
     if evaluator is None:
-        raise ValueError(f"Unknown evaluator: {evaluator_name}. Available evaluators: {', '.join(EVALUATORS.keys())}")
+        raise ValueError(
+            f"Unknown evaluator: {evaluator_name}. Available evaluators: {', '.join(EVALUATORS.keys())}"
+        )
     return evaluator
 
 
@@ -83,7 +87,9 @@ async def main(agent_name: str, dataset_path: str, evaluator_name: str):
     """
     The main function for the evaluation CLI.
     """
-    print(f"Starting evaluation with agent='{agent_name}', dataset='{dataset_path}', evaluator='{evaluator_name}'")
+    print(
+        f"Starting evaluation with agent='{agent_name}', dataset='{dataset_path}', evaluator='{evaluator_name}'"
+    )
 
     # 1. Get the agent and evaluator
     agent = get_agent(agent_name)
@@ -104,12 +110,15 @@ async def main(agent_name: str, dataset_path: str, evaluator_name: str):
         # If dataset with same name and examples exists, read it.
         # Otherwise, a new one is created.
         dataset = client.create_dataset(
-            dataset_name, description=f"Dataset for {agent_name} evaluation with {evaluator_name}."
+            dataset_name,
+            description=f"Dataset for {agent_name} evaluation with {evaluator_name}.",
         )
         for example in dataset_examples:
             client.create_example(
                 inputs={"user_input": example["user_input"]},
-                outputs={"output": example.get("expected_output")} if "expected_output" in example else None,
+                outputs={"output": example.get("expected_output")}
+                if "expected_output" in example
+                else None,
                 dataset_id=dataset.id,
             )
         print(f"Created and populated dataset '{dataset_name}' for this run.")
@@ -146,4 +155,10 @@ if __name__ == "__main__":
     )
     args = parser.parse_args()
 
-    asyncio.run(main(agent_name=args.agent, dataset_path=args.dataset, evaluator_name=args.evaluator))
+    asyncio.run(
+        main(
+            agent_name=args.agent,
+            dataset_path=args.dataset,
+            evaluator_name=args.evaluator,
+        )
+    )

{universal_mcp_agents-0.1.2 → universal_mcp_agents-0.1.4}/src/evals/utils.py

@@ -8,7 +8,9 @@ from evals.dataset import load_dataset
 load_dotenv()
 
 
-def upload_runs_to_dataset(project_name: str, dataset_name: str, dataset_description: str = ""):
+def upload_runs_to_dataset(
+    project_name: str, dataset_name: str, dataset_description: str = ""
+):
     """
     Uploads runs from a LangSmith project to a dataset.
 
@@ -36,7 +38,9 @@ def upload_runs_to_dataset(project_name: str, dataset_name: str, dataset_descrip
         )
         example_count += 1
 
-    print(f"✅ Successfully uploaded {example_count} runs from project '{project_name}' to dataset '{dataset_name}'.")
+    print(
+        f"✅ Successfully uploaded {example_count} runs from project '{project_name}' to dataset '{dataset_name}'."
+    )
 
 
 def upload_dataset_from_file(
@@ -71,7 +75,9 @@ def upload_dataset_from_file(
         outputs = {key: example[key] for key in output_keys if key in example}
         client.create_example(inputs=inputs, outputs=outputs, dataset_id=dataset.id)
 
-    print(f"✅ Successfully uploaded {len(examples)} examples from '{file_path}' to dataset '{dataset_name}'.")
+    print(
+        f"✅ Successfully uploaded {len(examples)} examples from '{file_path}' to dataset '{dataset_name}'."
+    )
 
 
 if __name__ == "__main__":
@@ -79,22 +85,48 @@ if __name__ == "__main__":
     subparsers = parser.add_subparsers(dest="command", required=True)
 
     # Sub-parser for uploading runs from a project
-    parser_runs = subparsers.add_parser("upload-runs", help="Upload runs from a project to a dataset.")
-    parser_runs.add_argument("--project-name", required=True, help="The LangSmith project name.")
-    parser_runs.add_argument("--dataset-name", required=True, help="The target dataset name.")
+    parser_runs = subparsers.add_parser(
+        "upload-runs", help="Upload runs from a project to a dataset."
+    )
     parser_runs.add_argument(
-        "--dataset-description", default="Dataset from project runs.", help="Description for the dataset."
+        "--project-name", required=True, help="The LangSmith project name."
+    )
+    parser_runs.add_argument(
+        "--dataset-name", required=True, help="The target dataset name."
+    )
+    parser_runs.add_argument(
+        "--dataset-description",
+        default="Dataset from project runs.",
+        help="Description for the dataset.",
     )
 
     # Sub-parser for uploading a dataset from a file
-    parser_file = subparsers.add_parser("upload-file", help="Upload a dataset from a local file.")
-    parser_file.add_argument("--file-path", required=True, help="Path to the local dataset file (CSV or JSONL).")
-    parser_file.add_argument("--dataset-name", required=True, help="The name for the dataset in LangSmith.")
+    parser_file = subparsers.add_parser(
+        "upload-file", help="Upload a dataset from a local file."
+    )
+    parser_file.add_argument(
+        "--file-path",
+        required=True,
+        help="Path to the local dataset file (CSV or JSONL).",
+    )
+    parser_file.add_argument(
+        "--dataset-name", required=True, help="The name for the dataset in LangSmith."
+    )
+    parser_file.add_argument(
+        "--dataset-description",
+        default="Dataset uploaded from file.",
+        help="Description for the dataset.",
+    )
+    parser_file.add_argument(
+        "--input-keys",
+        required=True,
+        help="Comma-separated list of input column names.",
+    )
     parser_file.add_argument(
-        "--dataset-description", default="Dataset uploaded from file.", help="Description for the dataset."
+        "--output-keys",
+        required=True,
+        help="Comma-separated list of output column names.",
     )
-    parser_file.add_argument("--input-keys", required=True, help="Comma-separated list of input column names.")
-    parser_file.add_argument("--output-keys", required=True, help="Comma-separated list of output column names.")
 
     args = parser.parse_args()
 

{universal_mcp_agents-0.1.2 → universal_mcp_agents-0.1.4}/src/tests/test_agents.py

@@ -21,14 +21,26 @@ class MockToolRegistry(ToolRegistry):
     def __init__(self, **kwargs: Any):
         """Initialize the MockToolRegistry."""
         self._apps = [
-            {"id": "google-mail", "name": "google-mail", "description": "Send and manage emails."},
-            {"id": "slack", "name": "slack", "description": "Team communication and messaging."},
+            {
+                "id": "google-mail",
+                "name": "google-mail",
+                "description": "Send and manage emails.",
+            },
+            {
+                "id": "slack",
+                "name": "slack",
+                "description": "Team communication and messaging.",
+            },
             {
                 "id": "google-calendar",
                 "name": "google-calendar",
                 "description": "Schedule and manage calendar events.",
             },
-            {"id": "jira", "name": "jira", "description": "Project tracking and issue management."},
+            {
+                "id": "jira",
+                "name": "jira",
+                "description": "Project tracking and issue management.",
+            },
             {
                 "id": "github",
                 "name": "github",
@@ -38,23 +50,67 @@ class MockToolRegistry(ToolRegistry):
         self._connected_apps = ["google-mail", "google-calendar", "github"]
         self._tools = {
             "google-mail": [
-                {"id": "send_email", "name": "send_email", "description": "Send an email to a recipient."},
-                {"id": "read_email", "name": "read_email", "description": "Read emails from inbox."},
-                {"id": "create_draft", "name": "create_draft", "description": "Create a draft email."},
+                {
+                    "id": "send_email",
+                    "name": "send_email",
+                    "description": "Send an email to a recipient.",
+                },
+                {
+                    "id": "read_email",
+                    "name": "read_email",
+                    "description": "Read emails from inbox.",
+                },
+                {
+                    "id": "create_draft",
+                    "name": "create_draft",
+                    "description": "Create a draft email.",
+                },
             ],
             "slack": [
-                {"id": "send_message", "name": "send_message", "description": "Send a message to a team channel."},
-                {"id": "read_channel", "name": "read_channel", "description": "Read messages from a channel."},
+                {
+                    "id": "send_message",
+                    "name": "send_message",
+                    "description": "Send a message to a team channel.",
+                },
+                {
+                    "id": "read_channel",
+                    "name": "read_channel",
+                    "description": "Read messages from a channel.",
+                },
             ],
             "google-calendar": [
-                {"id": "create_event", "name": "create_event", "description": "Create a new calendar event."},
-                {"id": "find_event", "name": "find_event", "description": "Find an event in the calendar."},
+                {
+                    "id": "create_event",
+                    "name": "create_event",
+                    "description": "Create a new calendar event.",
+                },
+                {
+                    "id": "find_event",
+                    "name": "find_event",
+                    "description": "Find an event in the calendar.",
+                },
             ],
             "github": [
-                {"id": "create_issue", "name": "create_issue", "description": "Create an issue in a repository."},
-                {"id": "get_issue", "name": "get_issue", "description": "Get details of a specific issue."},
-                {"id": "create_pull_request", "name": "create_pull_request", "description": "Create a pull request."},
-                {"id": "get_repository", "name": "get_repository", "description": "Get details of a repository."},
+                {
+                    "id": "create_issue",
+                    "name": "create_issue",
+                    "description": "Create an issue in a repository.",
+                },
+                {
+                    "id": "get_issue",
+                    "name": "get_issue",
+                    "description": "Get details of a specific issue.",
+                },
+                {
+                    "id": "create_pull_request",
+                    "name": "create_pull_request",
+                    "description": "Create a pull request.",
+                },
+                {
+                    "id": "get_repository",
+                    "name": "get_repository",
+                    "description": "Get details of a repository.",
+                },
             ],
         }
         self._tool_mappings = {
@@ -94,7 +150,11 @@ class MockToolRegistry(ToolRegistry):
     ) -> list[dict[str, Any]]:
         """Search for apps by a query."""
         query = query.lower()
-        results = [app for app in self._apps if query in app["name"].lower() or query in app["description"].lower()]
+        results = [
+            app
+            for app in self._apps
+            if query in app["name"].lower() or query in app["description"].lower()
+        ]
         return results[:limit]
 
     async def list_tools(
@@ -126,7 +186,12 @@ class MockToolRegistry(ToolRegistry):
         results = [tool for tool in all_app_tools if tool["name"] in found_tool_names]
 
         if not results:
-            results = [{"name": "general_purpose_tool", "description": "A general purpose tool."}]
+            results = [
+                {
+                    "name": "general_purpose_tool",
+                    "description": "A general purpose tool.",
+                }
+            ]
 
         return results[:limit]
 
@@ -145,7 +210,9 @@ class MockToolRegistry(ToolRegistry):
         # Return a list of mock tools for the ReAct agent to use
         return [mock_tool_callable]
 
-    async def call_tool(self, tool_name: str, tool_args: dict[str, Any]) -> dict[str, Any]:
+    async def call_tool(
+        self, tool_name: str, tool_args: dict[str, Any]
+    ) -> dict[str, Any]:
         """Call a tool with the given name and arguments."""
         print(f"MockToolRegistry: Called tool '{tool_name}' with args {tool_args}")
         return {"status": f"task has been done by tool {tool_name}"}
@@ -172,7 +239,9 @@ class TestToolFinderGraph:
         """Test Case 1: Simple case (Connected App)"""
         task = "Send an email to my manager about the project update."
         graph = build_tool_node_graph(llm, registry)
-        final_state = await graph.ainvoke({"task": task, "messages": [HumanMessage(content=task)]})
+        final_state = await graph.ainvoke(
+            {"task": task, "messages": [HumanMessage(content=task)]}
+        )
         assert final_state["apps_required"] is True
         assert "google-mail" in final_state["relevant_apps"]
         assert "google-mail" in final_state["apps_with_tools"]
@@ -183,7 +252,9 @@ class TestToolFinderGraph:
         """Test Case 2: Multiple apps found"""
         task = "Send a message to my team about the new design."
         graph = build_tool_node_graph(llm, registry)
-        final_state = await graph.ainvoke({"task": task, "messages": [HumanMessage(content=task)]})
+        final_state = await graph.ainvoke(
+            {"task": task, "messages": [HumanMessage(content=task)]}
+        )
         assert final_state["apps_required"] is True
         assert "google-mail" in final_state["relevant_apps"]
         assert "slack" in final_state["relevant_apps"]
@@ -195,7 +266,9 @@ class TestToolFinderGraph:
         """Test Case 3: No relevant app"""
         task = "Can you create a blog post on my wordpress site?"
         graph = build_tool_node_graph(llm, registry)
-        final_state = await graph.ainvoke({"task": task, "messages": [HumanMessage(content=task)]})
+        final_state = await graph.ainvoke(
+            {"task": task, "messages": [HumanMessage(content=task)]}
+        )
         assert final_state["apps_required"] is True
         assert not final_state["relevant_apps"]
         assert not final_state["apps_with_tools"]
@@ -205,7 +278,9 @@ class TestToolFinderGraph:
         """Test Case 4: Multiple tools in one app"""
         task = "Create a new issue for a bug in our github repository, and send message on slack about the issue."
         graph = build_tool_node_graph(llm, registry)
-        final_state = await graph.ainvoke({"task": task, "messages": [HumanMessage(content=task)]})
+        final_state = await graph.ainvoke(
+            {"task": task, "messages": [HumanMessage(content=task)]}
+        )
         assert final_state["apps_required"] is True
         assert "github" in final_state["relevant_apps"]
         assert "slack" in final_state["relevant_apps"]
@@ -219,7 +294,9 @@ class TestToolFinderGraph:
         """Test Case 5: Unavailable App"""
         task = "Create a new design file in Figma."
         graph = build_tool_node_graph(llm, registry)
-        final_state = await graph.ainvoke({"task": task, "messages": [HumanMessage(content=task)]})
+        final_state = await graph.ainvoke(
+            {"task": task, "messages": [HumanMessage(content=task)]}
+        )
         assert final_state["apps_required"] is True
         assert not final_state["relevant_apps"]
         assert not final_state["apps_with_tools"]
@@ -229,7 +306,9 @@ class TestToolFinderGraph:
         """Test Case 6: No App Needed"""
         task = "hello"
         graph = build_tool_node_graph(llm, registry)
-        final_state = await graph.ainvoke({"task": task, "messages": [HumanMessage(content=task)]})
+        final_state = await graph.ainvoke(
+            {"task": task, "messages": [HumanMessage(content=task)]}
+        )
         assert final_state["apps_required"] is False
 
 
@@ -272,7 +351,11 @@ class TestAgents:
         assert final_messages, "The agent should have produced at least one message."
         last_message = final_messages[-1]
 
-        final_response = last_message.content if hasattr(last_message, "content") else str(last_message)
+        final_response = (
+            last_message.content
+            if hasattr(last_message, "content")
+            else str(last_message)
+        )
 
         # Print the response for manual verification and for the LLM judge
         print("\n--- Agent's Final Response ---")

{universal_mcp_agents-0.1.2 → universal_mcp_agents-0.1.4}/src/universal_mcp/agents/autoagent/graph.py

@@ -22,7 +22,9 @@ async def build_graph(tool_registry: ToolRegistry, instructions: str = ""):
         tools_list = []
         if app_ids is not None:
             for app_id in app_ids:
-                tools_list.extend(await tool_registry.search_tools(query, limit=10, app_id=app_id))
+                tools_list.extend(
+                    await tool_registry.search_tools(query, limit=10, app_id=app_id)
+                )
         else:
             tools_list = await tool_registry.search_tools(query, limit=10)
         tools_list = [f"{tool['id']}: {tool['description']}" for tool in tools_list]
@@ -48,21 +50,33 @@ async def build_graph(tool_registry: ToolRegistry, instructions: str = ""):
         connections = await tool_registry.list_connected_apps()
         connection_ids = set([connection["app_id"] for connection in connections])
         connected_apps = [app["id"] for app in app_ids if app["id"] in connection_ids]
-        unconnected_apps = [app["id"] for app in app_ids if app["id"] not in connection_ids]
-        app_id_descriptions = "These are the apps connected to the user's account:\n" + "\n".join(
-            [f"{app}" for app in connected_apps]
+        unconnected_apps = [
+            app["id"] for app in app_ids if app["id"] not in connection_ids
+        ]
+        app_id_descriptions = (
+            "These are the apps connected to the user's account:\n"
+            + "\n".join([f"{app}" for app in connected_apps])
         )
         if unconnected_apps:
             app_id_descriptions += "\n\nOther (not connected) apps: " + "\n".join(
                 [f"{app}" for app in unconnected_apps]
             )
 
-        system_prompt = system_prompt.format(system_time=datetime.now(tz=UTC).isoformat(), app_ids=app_id_descriptions)
+        system_prompt = system_prompt.format(
+            system_time=datetime.now(tz=UTC).isoformat(), app_ids=app_id_descriptions
+        )
 
-        messages = [{"role": "system", "content": system_prompt + "\n" + instructions}, *state["messages"]]
+        messages = [
+            {"role": "system", "content": system_prompt + "\n" + instructions},
+            *state["messages"],
+        ]
         model = load_chat_model(runtime.context.model)
-        loaded_tools = await tool_registry.export_tools(tools=state["selected_tool_ids"], format=ToolFormat.LANGCHAIN)
-        model_with_tools = model.bind_tools([search_tools, ask_user, load_tools, *loaded_tools], tool_choice="auto")
+        loaded_tools = await tool_registry.export_tools(
+            tools=state["selected_tool_ids"], format=ToolFormat.LANGCHAIN
+        )
+        model_with_tools = model.bind_tools(
+            [search_tools, ask_user, load_tools, *loaded_tools], tool_choice="auto"
+        )
         response_raw = model_with_tools.invoke(messages)
         response = cast(AIMessage, response_raw)
         return {"messages": [response]}
@@ -102,7 +116,8 @@ async def build_graph(tool_registry: ToolRegistry, instructions: str = ""):
                 tools = await search_tools.ainvoke(tool_call["args"])
                 outputs.append(
                     ToolMessage(
-                        content=json.dumps(tools) + "\n\nUse the load_tools tool to load the tools you want to use.",
+                        content=json.dumps(tools)
+                        + "\n\nUse the load_tools tool to load the tools you want to use.",
                         name=tool_call["name"],
                         tool_call_id=tool_call["id"],
                     )
@@ -119,9 +134,13 @@ async def build_graph(tool_registry: ToolRegistry, instructions: str = ""):
                     )
                 )
             else:
-                await tool_registry.export_tools([tool_call["name"]], ToolFormat.LANGCHAIN)
+                await tool_registry.export_tools(
+                    [tool_call["name"]], ToolFormat.LANGCHAIN
+                )
                 try:
-                    tool_result = await tool_registry.call_tool(tool_call["name"], tool_call["args"])
+                    tool_result = await tool_registry.call_tool(
+                        tool_call["name"], tool_call["args"]
+                    )
                     outputs.append(
                         ToolMessage(
                             content=json.dumps(tool_result),

{universal_mcp_agents-0.1.2 → universal_mcp_agents-0.1.4}/src/universal_mcp/agents/autoagent/studio.py

@@ -8,7 +8,6 @@ tool_registry = AgentrRegistry()
 tool_manager = ToolManager()
 
 
-
 async def main():
     instructions = """
     You are a helpful assistant that can use tools to help the user. If a task requires multiple steps, you should perform separate different searches for different actions. Prefer completing one action before searching for another.
@@ -16,10 +15,5 @@ async def main():
     graph = await build_graph(tool_registry, instructions=instructions)
     return graph
 
-graph = asyncio.run(main())
-
-
-
-
-
 
+graph = asyncio.run(main())

{universal_mcp_agents-0.1.2 → universal_mcp_agents-0.1.4}/src/universal_mcp/agents/base.py

@@ -10,7 +10,14 @@ from .utils import RichCLI
 
 
 class BaseAgent:
-    def __init__(self, name: str, instructions: str, model: str, memory: BaseCheckpointSaver | None = None, **kwargs):
+    def __init__(
+        self,
+        name: str,
+        instructions: str,
+        model: str,
+        memory: BaseCheckpointSaver | None = None,
+        **kwargs,
+    ):
         self.name = name
         self.instructions = instructions
         self.model = model
@@ -27,12 +34,26 @@ class BaseAgent:
     async def _build_graph(self):
         raise NotImplementedError("Subclasses must implement this method")
 
-    async def stream(self, thread_id: str, user_input: str):
+    async def stream(self, thread_id: str, user_input: str, metadata: dict = None):
         await self.ainit()
         aggregate = None
+
+        run_metadata = {
+            "agent_name": self.name,
+            "is_background_run": False,  # Default to False
+        }
+
+        if metadata:
+            run_metadata.update(metadata)
+
+        run_config = {
+            "configurable": {"thread_id": thread_id},
+            "metadata": run_metadata,
+        }
+
         async for event, metadata in self._graph.astream(
             {"messages": [{"role": "user", "content": user_input}]},
-            config={"configurable": {"thread_id": thread_id}},
+            config=run_config,
             context={"system_prompt": self.instructions, "model": self.model},
             stream_mode="messages",
             stream_usage=True,
@@ -66,12 +87,28 @@ class BaseAgent:
             async for event in self.stream(thread_id, user_input):
                 stream_updater.update(event.content)
 
-    async def invoke(self, user_input: str, thread_id: str = str(uuid4())):
+    async def invoke(
+        self, user_input: str, thread_id: str = str(uuid4()), metadata: dict = None
+    ):
         """Run the agent"""
         await self.ainit()
+
+        run_metadata = {
+            "agent_name": self.name,
+            "is_background_run": False,  # Default to False
+        }
+
+        if metadata:
+            run_metadata.update(metadata)
+
+        run_config = {
+            "configurable": {"thread_id": thread_id},
+            "metadata": run_metadata,
+        }
+
         return await self._graph.ainvoke(
             {"messages": [{"role": "user", "content": user_input}]},
-            config={"configurable": {"thread_id": thread_id}},
+            config=run_config,
             context={"system_prompt": self.instructions, "model": self.model},
         )
 
@@ -85,10 +122,15 @@ class BaseAgent:
         # Main loop
         while True:
             try:
-                state = self._graph.get_state(config={"configurable": {"thread_id": thread_id}})
+                state = self._graph.get_state(
+                    config={"configurable": {"thread_id": thread_id}}
+                )
                 if state.interrupts:
                     value = self.cli.handle_interrupt(state.interrupts[0])
-                    self._graph.invoke(Command(resume=value), config={"configurable": {"thread_id": thread_id}})
+                    self._graph.invoke(
+                        Command(resume=value),
+                        config={"configurable": {"thread_id": thread_id}},
+                    )
                     continue
 
                 user_input = self.cli.get_user_input()
@@ -99,7 +141,9 @@ class BaseAgent:
                 if user_input.startswith("/"):
                     command = user_input.lower().lstrip("/")
                     if command == "about":
-                        self.cli.display_info(f"Agent is {self.name}. {self.instructions}")
+                        self.cli.display_info(
+                            f"Agent is {self.name}. {self.instructions}"
+                        )
                         continue
                     elif command == "exit" or command == "quit" or command == "q":
                         self.cli.display_info("Goodbye! 👋")
@@ -110,7 +154,9 @@ class BaseAgent:
                         thread_id = str(uuid4())
                         continue
                     elif command == "help":
-                        self.cli.display_info("Available commands: /about, /exit, /quit, /q, /reset")
+                        self.cli.display_info(
+                            "Available commands: /about, /exit, /quit, /q, /reset"
+                        )
                         continue
                     else:
                         self.cli.display_error(f"Unknown command: {command}")