fast-agent-mcp 0.1.10__tar.gz → 0.1.11__tar.gz

{fast_agent_mcp-0.1.10 → fast_agent_mcp-0.1.11}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: fast-agent-mcp
-Version: 0.1.10
+Version: 0.1.11
 Summary: Define, Prompt and Test MCP enabled Agents and Workflows
 Author-email: Shaun Smith <fastagent@llmindset.co.uk>, Sarmad Qadri <sarmad@lastmileai.dev>
 License:                                  Apache License
@@ -212,7 +212,7 @@ Requires-Python: >=3.10
 Requires-Dist: aiohttp>=3.11.13
 Requires-Dist: anthropic>=0.49.0
 Requires-Dist: fastapi>=0.115.6
-Requires-Dist: mcp>=1.4.1
+Requires-Dist: mcp>=1.5.0
 Requires-Dist: numpy>=2.2.1
 Requires-Dist: openai>=1.63.2
 Requires-Dist: opentelemetry-distro>=0.50b0
@@ -241,10 +241,9 @@ Provides-Extra: temporal
 Requires-Dist: temporalio>=1.8.0; extra == 'temporal'
 Description-Content-Type: text/markdown
 
-## fast-agent
-
 <p align="center">
 <a href="https://pypi.org/project/fast-agent-mcp/"><img src="https://img.shields.io/pypi/v/fast-agent-mcp?color=%2334D058&label=pypi" /></a>
+<a href="#"><img src="https://github.com/evalstate/fast-agent/actions/workflows/main-checks.yml/badge.svg" /></a>
 <a href="https://github.com/evalstate/fast-agent/issues"><img src="https://img.shields.io/github/issues-raw/evalstate/fast-agent" /></a>
 <a href="https://lmai.link/discord/mcp-agent"><img src="https://shields.io/discord/1089284610329952357" alt="discord" /></a>
 <img alt="Pepy Total Downloads" src="https://img.shields.io/pepy/dt/fast-agent-mcp?label=pypi%20%7C%20downloads"/>
@@ -253,16 +252,14 @@ Description-Content-Type: text/markdown
 
 ## Overview
 
-**`fast-agent`** enables you to create and interact with sophisticated Agents and Workflows in minutes.
+**`fast-agent`** enables you to create and interact with sophisticated Agents and Workflows in minutes. It is the first framework with complete, end-to-end tested MCP Feature support including Sampling. Both Anthropic (Haiku, Sonnet, Opus) and OpenAI models (gpt-4o family, o1/o3 family) are supported.
 
 The simple declarative syntax lets you concentrate on composing your Prompts and MCP Servers to [build effective agents](https://www.anthropic.com/research/building-effective-agents).
 
-Evaluate how different models handle Agent and MCP Server calling tasks, then build multi-model workflows using the best provider for each task.
-
-`fast-agent` is now multi-modal, supporting Images and PDFs for both Anthropic and OpenAI endpoints (for supported models), via Prompts, Resources and MCP Tool Call results.
+`fast-agent` is multi-modal, supporting Images and PDFs for both Anthropic and OpenAI endpoints via Prompts, Resources and MCP Tool Call results. The inclusion of passthrough and playback LLMs enable rapid development and test of Python glue-code for your applications.
 
 > [!TIP]
-> `fast-agent` is now MCP Native! Coming Soon - Full Documentation Site.
+> `fast-agent` is now MCP Native! Coming Soon - Full Documentation Site and Further MCP Examples.
 
 ### Agent Application Development
 
@@ -272,7 +269,7 @@ Chat with individual Agents and Components before, during and after workflow exe
 
 Simple model selection makes testing Model <-> MCP Server interaction painless. You can read more about the motivation behind this project [here](https://llmindset.co.uk/resources/fast-agent/)
 
-![fast-agent](https://github.com/user-attachments/assets/3e692103-bf97-489a-b519-2d0fee036369)
+![2025-03-23-fast-agent](https://github.com/user-attachments/assets/8f6dbb69-43e3-4633-8e12-5572e9614728)
 
 ## Get started:
 
@@ -597,6 +594,14 @@ agent["greeter"].send("Good Evening!")          # Dictionary access is supported
 
 Add Resources to prompts using either the inbuilt `prompt-server` or MCP Types directly. Convenience class are made available to do so simply, for example:
 
+```python
+  summary: str =  await agent.with_resource(
+      "Summarise this PDF please",
+      "mcp_server",
+      "resource://fast-agent/sample.pdf",
+  )
+```
+
 #### MCP Tool Result Conversion
 
 LLM APIs have restrictions on the content types that can be returned as Tool Calls/Function results via their Chat Completions API's:
@@ -612,40 +617,33 @@ MCP Prompts are supported with `apply_prompt(name,arguments)`, which always retu
 
 Prompts can also be applied interactively through the interactive interface by using the `/prompt` command.
 
+### Sampling
+
+Sampling LLMs are configured per Client/Server pair. Specify the model name in fastagent.config.yaml as follows:
+
+```yaml
+mcp:
+  servers:
+    sampling_resource:
+      command: "uv"
+      args: ["run", "sampling_resource_server.py"]
+      sampling:
+        model: "haiku"
+```
+
 ### Secrets File
 
 > [!TIP]
 > fast-agent will look recursively for a fastagent.secrets.yaml file, so you only need to manage this at the root folder of your agent definitions.
 
+### Interactive Shell
+
+![fast-agent](https://github.com/user-attachments/assets/3e692103-bf97-489a-b519-2d0fee036369)
+
 ## Project Notes
 
 `fast-agent` builds on the [`mcp-agent`](https://github.com/lastmile-ai/mcp-agent) project by Sarmad Qadri.
 
-### llmindset.co.uk fork:
-
-- Addition of MCP Prompts including Prompt Server and agent save/replay ability.
-- Overhaul of Eval/Opt for Conversation Management
-- Removed instructor/double-llm calling - native structured outputs for OAI.
-- Improved handling of Parallel/Fan-In and respose option
-- XML based generated prompts
-- "FastAgent" style prototyping, with per-agent models
-- API keys through Environment Variables
-- Warm-up / Post-Workflow Agent Interactions
-- Quick Setup
-- Interactive Prompt Mode
-- Simple Model Selection with aliases
-- User/Assistant and Tool Call message display
-- MCP Sever Environment Variable support
-- MCP Roots support
-- Comprehensive Progress display
-- JSONL file logging with secret revokation
-- OpenAI o1/o3-mini support with reasoning level
-- Enhanced Human Input Messaging and Handling
-- Declarative workflows
-- Numerous defect fixes
-
-### Features to add (Commmitted)
-
-- Run Agent as MCP Server, with interop
-- Multi-part content types supporing Vision, PDF and multi-part Text.
-- Improved test automation (supported by prompt_server.py and augmented_llm_playback.py)
+### Contributing
+
+Contributions and PRs are welcome - feel free to raise issues to discuss. Full guidelines for contributing and roadmap coming very soon. Get in touch!

{fast_agent_mcp-0.1.10 → fast_agent_mcp-0.1.11}/README.md

@@ -1,7 +1,6 @@
-## fast-agent
-
 <p align="center">
 <a href="https://pypi.org/project/fast-agent-mcp/"><img src="https://img.shields.io/pypi/v/fast-agent-mcp?color=%2334D058&label=pypi" /></a>
+<a href="#"><img src="https://github.com/evalstate/fast-agent/actions/workflows/main-checks.yml/badge.svg" /></a>
 <a href="https://github.com/evalstate/fast-agent/issues"><img src="https://img.shields.io/github/issues-raw/evalstate/fast-agent" /></a>
 <a href="https://lmai.link/discord/mcp-agent"><img src="https://shields.io/discord/1089284610329952357" alt="discord" /></a>
 <img alt="Pepy Total Downloads" src="https://img.shields.io/pepy/dt/fast-agent-mcp?label=pypi%20%7C%20downloads"/>
@@ -10,16 +9,14 @@
 
 ## Overview
 
-**`fast-agent`** enables you to create and interact with sophisticated Agents and Workflows in minutes.
+**`fast-agent`** enables you to create and interact with sophisticated Agents and Workflows in minutes. It is the first framework with complete, end-to-end tested MCP Feature support including Sampling. Both Anthropic (Haiku, Sonnet, Opus) and OpenAI models (gpt-4o family, o1/o3 family) are supported.
 
 The simple declarative syntax lets you concentrate on composing your Prompts and MCP Servers to [build effective agents](https://www.anthropic.com/research/building-effective-agents).
 
-Evaluate how different models handle Agent and MCP Server calling tasks, then build multi-model workflows using the best provider for each task.
-
-`fast-agent` is now multi-modal, supporting Images and PDFs for both Anthropic and OpenAI endpoints (for supported models), via Prompts, Resources and MCP Tool Call results.
+`fast-agent` is multi-modal, supporting Images and PDFs for both Anthropic and OpenAI endpoints via Prompts, Resources and MCP Tool Call results. The inclusion of passthrough and playback LLMs enable rapid development and test of Python glue-code for your applications.
 
 > [!TIP]
-> `fast-agent` is now MCP Native! Coming Soon - Full Documentation Site.
+> `fast-agent` is now MCP Native! Coming Soon - Full Documentation Site and Further MCP Examples.
 
 ### Agent Application Development
 
@@ -29,7 +26,7 @@ Chat with individual Agents and Components before, during and after workflow exe
 
 Simple model selection makes testing Model <-> MCP Server interaction painless. You can read more about the motivation behind this project [here](https://llmindset.co.uk/resources/fast-agent/)
 
-![fast-agent](https://github.com/user-attachments/assets/3e692103-bf97-489a-b519-2d0fee036369)
+![2025-03-23-fast-agent](https://github.com/user-attachments/assets/8f6dbb69-43e3-4633-8e12-5572e9614728)
 
 ## Get started:
 
@@ -354,6 +351,14 @@ agent["greeter"].send("Good Evening!")          # Dictionary access is supported
 
 Add Resources to prompts using either the inbuilt `prompt-server` or MCP Types directly. Convenience class are made available to do so simply, for example:
 
+```python
+  summary: str =  await agent.with_resource(
+      "Summarise this PDF please",
+      "mcp_server",
+      "resource://fast-agent/sample.pdf",
+  )
+```
+
 #### MCP Tool Result Conversion
 
 LLM APIs have restrictions on the content types that can be returned as Tool Calls/Function results via their Chat Completions API's:
@@ -369,40 +374,33 @@ MCP Prompts are supported with `apply_prompt(name,arguments)`, which always retu
 
 Prompts can also be applied interactively through the interactive interface by using the `/prompt` command.
 
+### Sampling
+
+Sampling LLMs are configured per Client/Server pair. Specify the model name in fastagent.config.yaml as follows:
+
+```yaml
+mcp:
+  servers:
+    sampling_resource:
+      command: "uv"
+      args: ["run", "sampling_resource_server.py"]
+      sampling:
+        model: "haiku"
+```
+
 ### Secrets File
 
 > [!TIP]
 > fast-agent will look recursively for a fastagent.secrets.yaml file, so you only need to manage this at the root folder of your agent definitions.
 
+### Interactive Shell
+
+![fast-agent](https://github.com/user-attachments/assets/3e692103-bf97-489a-b519-2d0fee036369)
+
 ## Project Notes
 
 `fast-agent` builds on the [`mcp-agent`](https://github.com/lastmile-ai/mcp-agent) project by Sarmad Qadri.
 
-### llmindset.co.uk fork:
-
-- Addition of MCP Prompts including Prompt Server and agent save/replay ability.
-- Overhaul of Eval/Opt for Conversation Management
-- Removed instructor/double-llm calling - native structured outputs for OAI.
-- Improved handling of Parallel/Fan-In and respose option
-- XML based generated prompts
-- "FastAgent" style prototyping, with per-agent models
-- API keys through Environment Variables
-- Warm-up / Post-Workflow Agent Interactions
-- Quick Setup
-- Interactive Prompt Mode
-- Simple Model Selection with aliases
-- User/Assistant and Tool Call message display
-- MCP Sever Environment Variable support
-- MCP Roots support
-- Comprehensive Progress display
-- JSONL file logging with secret revokation
-- OpenAI o1/o3-mini support with reasoning level
-- Enhanced Human Input Messaging and Handling
-- Declarative workflows
-- Numerous defect fixes
-
-### Features to add (Commmitted)
-
-- Run Agent as MCP Server, with interop
-- Multi-part content types supporing Vision, PDF and multi-part Text.
-- Improved test automation (supported by prompt_server.py and augmented_llm_playback.py)
+### Contributing
+
+Contributions and PRs are welcome - feel free to raise issues to discuss. Full guidelines for contributing and roadmap coming very soon. Get in touch!

{fast_agent_mcp-0.1.10 → fast_agent_mcp-0.1.11}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "fast-agent-mcp"
-version = "0.1.10"
+version = "0.1.11"
 description = "Define, Prompt and Test MCP enabled Agents and Workflows"
 readme = "README.md"
 license = { file = "LICENSE" }
@@ -16,7 +16,7 @@ classifiers = [
 requires-python = ">=3.10"
 dependencies = [
     "fastapi>=0.115.6",
-    "mcp>=1.4.1",
+    "mcp>=1.5.0",
     "opentelemetry-distro>=0.50b0",
     "opentelemetry-exporter-otlp-proto-http>=1.29.0",
     "pydantic-settings>=2.7.0",

{fast_agent_mcp-0.1.10 → fast_agent_mcp-0.1.11}/src/mcp_agent/config.py

@@ -18,6 +18,12 @@ class MCPServerAuthSettings(BaseModel):
     model_config = ConfigDict(extra="allow", arbitrary_types_allowed=True)
 
 
+class MCPSamplingSettings(BaseModel):
+    model: str = "haiku"
+
+    model_config = ConfigDict(extra="allow", arbitrary_types_allowed=True)
+
+
 class MCPRootSettings(BaseModel):
     """Represents a root directory configuration for an MCP server."""
 
@@ -81,6 +87,9 @@ class MCPServerSettings(BaseModel):
     env: Dict[str, str] | None = None
     """Environment variables to pass to the server process."""
 
+    sampling: MCPSamplingSettings | None = None
+    """Sampling settings for this Client/Server pair"""
+
 
 class MCPSettings(BaseModel):
     """Configuration for all MCP servers."""

{fast_agent_mcp-0.1.10 → fast_agent_mcp-0.1.11}/src/mcp_agent/core/agent_app.py

@@ -112,6 +112,35 @@ class AgentApp:
 
         proxy = self._agents[target]
         return await proxy.apply_prompt(prompt_name, arguments)
+        
+    async def with_resource(
+        self,
+        prompt_content: Union[str, PromptMessageMultipart],
+        server_name: str,
+        resource_name: str,
+        agent_name: Optional[str] = None,
+    ) -> str:
+        """
+        Create a prompt with the given content and resource, then send it to the agent.
+
+        Args:
+            prompt_content: Either a string message or an existing PromptMessageMultipart
+            server_name: Name of the MCP server to retrieve the resource from
+            resource_name: Name or URI of the resource to retrieve
+            agent_name: The name of the agent to use (uses default if None)
+
+        Returns:
+            The agent's response as a string
+        """
+        target = agent_name or self._default
+        if not target:
+            raise ValueError("No default agent available")
+
+        if target not in self._agents:
+            raise ValueError(f"No agent named '{target}'")
+
+        proxy = self._agents[target]
+        return await proxy.with_resource(prompt_content, server_name, resource_name)
 
     async def prompt(self, agent_name: Optional[str] = None, default: str = "") -> str:
         """

{fast_agent_mcp-0.1.10 → fast_agent_mcp-0.1.11}/src/mcp_agent/core/proxies.py

@@ -1,6 +1,9 @@
 """
 Proxy classes for agent interactions.
 These proxies provide a consistent interface for interacting with different types of agents.
+
+FOR COMPATIBILITY WITH LEGACY MCP-AGENT CODE
+
 """
 
 from typing import List, Optional, Dict, Union, TYPE_CHECKING

{fast_agent_mcp-0.1.10 → fast_agent_mcp-0.1.11}/src/mcp_agent/mcp/gen_client.py

@@ -6,7 +6,7 @@ from anyio.streams.memory import MemoryObjectReceiveStream, MemoryObjectSendStre
 from mcp import ClientSession
 
 from mcp_agent.logging.logger import get_logger
-from mcp_agent.mcp_server_registry import ServerRegistry
+from mcp_agent.mcp.interfaces import ServerRegistryProtocol
 from mcp_agent.mcp.mcp_agent_client_session import MCPAgentClientSession
 
 logger = get_logger(__name__)
@@ -15,7 +15,7 @@ logger = get_logger(__name__)
 @asynccontextmanager
 async def gen_client(
     server_name: str,
-    server_registry: ServerRegistry,
+    server_registry: ServerRegistryProtocol,
     client_session_factory: Callable[
         [MemoryObjectReceiveStream, MemoryObjectSendStream, timedelta | None],
         ClientSession,
@@ -41,7 +41,7 @@ async def gen_client(
 
 async def connect(
     server_name: str,
-    server_registry: ServerRegistry,
+    server_registry: ServerRegistryProtocol,
     client_session_factory: Callable[
         [MemoryObjectReceiveStream, MemoryObjectSendStream, timedelta | None],
         ClientSession,
@@ -67,7 +67,7 @@ async def connect(
 
 async def disconnect(
     server_name: str | None,
-    server_registry: ServerRegistry,
+    server_registry: ServerRegistryProtocol,
 ) -> None:
     """
     Disconnect from the specified server. If server_name is None, disconnect from all servers.

fast_agent_mcp-0.1.11/src/mcp_agent/mcp/interfaces.py

@@ -0,0 +1,152 @@
+"""
+Interface definitions to prevent circular imports.
+This module defines protocols (interfaces) that can be used to break circular dependencies.
+"""
+
+from contextlib import asynccontextmanager
+from typing import Any, AsyncGenerator, Callable, Generic, List, Optional, Protocol, Type, TypeVar
+
+from mcp import ClientSession
+from mcp.types import CreateMessageRequestParams
+from pydantic import Field
+
+
+class ServerRegistryProtocol(Protocol):
+    """
+    Protocol defining the minimal interface of ServerRegistry needed by gen_client.
+    This allows gen_client to depend on this protocol rather than the full ServerRegistry class.
+    """
+
+    @asynccontextmanager
+    async def initialize_server(
+        self,
+        server_name: str,
+        client_session_factory=None,
+        init_hook=None,
+    ) -> AsyncGenerator[ClientSession, None]:
+        """Initialize a server and yield a client session."""
+        ...
+
+    @property
+    def connection_manager(self) -> "ConnectionManagerProtocol":
+        """Get the connection manager."""
+        ...
+
+
+class ConnectionManagerProtocol(Protocol):
+    """
+    Protocol defining the minimal interface of ConnectionManager needed.
+    """
+
+    async def get_server(
+        self,
+        server_name: str,
+        client_session_factory=None,
+    ):
+        """Get a server connection."""
+        ...
+
+    async def disconnect_server(self, server_name: str) -> None:
+        """Disconnect from a server."""
+        ...
+
+    async def disconnect_all_servers(self) -> None:
+        """Disconnect from all servers."""
+        ...
+
+
+# Type variables for generic protocols
+MessageParamT = TypeVar("MessageParamT")
+"""A type representing an input message to an LLM."""
+
+MessageT = TypeVar("MessageT")
+"""A type representing an output message from an LLM."""
+
+ModelT = TypeVar("ModelT")
+"""A type representing a structured output message from an LLM."""
+
+
+class RequestParams(CreateMessageRequestParams):
+    """
+    Parameters to configure the AugmentedLLM 'generate' requests.
+    """
+
+    messages: None = Field(exclude=True, default=None)
+    """
+    Ignored. 'messages' are removed from CreateMessageRequestParams 
+    to avoid confusion with the 'message' parameter on 'generate' method.
+    """
+
+    maxTokens: int = 2048
+    """The maximum number of tokens to sample, as requested by the server."""
+
+    model: str | None = None
+    """
+    The model to use for the LLM generation.
+    If specified, this overrides the 'modelPreferences' selection criteria.
+    """
+
+    use_history: bool = True
+    """
+    Include the message history in the generate request.
+    """
+
+    max_iterations: int = 10
+    """
+    The maximum number of iterations to run the LLM for.
+    """
+
+    parallel_tool_calls: bool = True
+    """
+    Whether to allow multiple tool calls per iteration.
+    Also known as multi-step tool use.
+    """
+
+
+class AugmentedLLMProtocol(Protocol, Generic[MessageParamT, MessageT]):
+    """Protocol defining the interface for augmented LLMs"""
+
+    async def generate(
+        self,
+        message: str | MessageParamT | List[MessageParamT],
+        request_params: RequestParams | None = None,
+    ) -> List[MessageT]:
+        """Request an LLM generation, which may run multiple iterations, and return the result"""
+
+    async def generate_str(
+        self,
+        message: str | MessageParamT | List[MessageParamT],
+        request_params: RequestParams | None = None,
+    ) -> str:
+        """Request an LLM generation and return the string representation of the result"""
+
+    async def generate_structured(
+        self,
+        message: str | MessageParamT | List[MessageParamT],
+        response_model: Type[ModelT],
+        request_params: RequestParams | None = None,
+    ) -> ModelT:
+        """Request a structured LLM generation and return the result as a Pydantic model."""
+
+
+class ModelFactoryClassProtocol(Protocol):
+    """
+    Protocol defining the minimal interface of the ModelFactory class needed by sampling.
+    This allows sampling.py to depend on this protocol rather than the concrete ModelFactory class.
+    """
+    
+    @classmethod
+    def create_factory(
+        cls, model_string: str, request_params: Optional[RequestParams] = None
+    ) -> Callable[..., AugmentedLLMProtocol[Any, Any]]:
+        """
+        Creates a factory function that can be used to construct an LLM instance.
+
+        Args:
+            model_string: The model specification string
+            request_params: Optional parameters to configure LLM behavior
+
+        Returns:
+            A factory function that can create an LLM instance
+        """
+        ...

{fast_agent_mcp-0.1.10 → fast_agent_mcp-0.1.11}/src/mcp_agent/mcp/mcp_agent_client_session.py

@@ -24,6 +24,7 @@ from pydantic import AnyUrl
 from mcp_agent.config import MCPServerSettings
 from mcp_agent.context_dependent import ContextDependent
 from mcp_agent.logging.logger import get_logger
+from mcp_agent.mcp.sampling import sample
 
 logger = get_logger(__name__)
 
@@ -40,7 +41,12 @@ async def list_roots(ctx: ClientSession) -> ListRootsResult:
         and ctx.session.server_config.roots
     ):
         roots = [
-            Root(uri=AnyUrl(root.uri), name=root.name)
+            Root(
+                uri=AnyUrl(
+                    root.server_uri_alias or root.uri,
+                ),
+                name=root.name,
+            )
             for root in ctx.session.server_config.roots
         ]
     return ListRootsResult(roots=roots or [])
@@ -58,7 +64,9 @@ class MCPAgentClientSession(ClientSession, ContextDependent):
     """
 
     def __init__(self, *args, **kwargs):
-        super().__init__(*args, **kwargs, list_roots_callback=list_roots)
+        super().__init__(
+            *args, **kwargs, list_roots_callback=list_roots, sampling_callback=sample
+        )
         self.server_config: Optional[MCPServerSettings] = None
 
     async def send_request(
@@ -115,4 +123,4 @@ class MCPAgentClientSession(ClientSession, ContextDependent):
         )
         return await super().send_progress_notification(
             progress_token=progress_token, progress=progress, total=total
-        )
+        )

{fast_agent_mcp-0.1.10 → fast_agent_mcp-0.1.11}/src/mcp_agent/mcp/mcp_aggregator.py

@@ -16,6 +16,7 @@ from mcp.server.stdio import stdio_server
 from mcp.types import (
     CallToolResult,
     ListToolsResult,
+    TextContent,
     Tool,
     Prompt,
 )
@@ -459,7 +460,10 @@ class MCPAggregator(ContextDependent):
 
         if server_name is None or local_tool_name is None:
             logger.error(f"Error: Tool '{name}' not found")
-            return CallToolResult(isError=True, message=f"Tool '{name}' not found")
+            return CallToolResult(
+                isError=True, 
+                content=[TextContent(type="text", text=f"Tool '{name}' not found")]
+            )
 
         logger.info(
             "Requesting tool call",
@@ -477,7 +481,10 @@ class MCPAggregator(ContextDependent):
             operation_name=local_tool_name,
             method_name="call_tool",
             method_args={"name": local_tool_name, "arguments": arguments},
-            error_factory=lambda msg: CallToolResult(isError=True, message=msg),
+            error_factory=lambda msg: CallToolResult(
+                isError=True, 
+                content=[TextContent(type="text", text=msg)]
+            ),
         )
 
     async def get_prompt(
@@ -898,7 +905,10 @@ class MCPCompoundServer(Server):
             result = await self.aggregator.call_tool(name=name, arguments=arguments)
             return result.content
         except Exception as e:
-            return CallToolResult(isError=True, message=f"Error calling tool: {e}")
+            return CallToolResult(
+                isError=True, 
+                content=[TextContent(type="text", text=f"Error calling tool: {e}")]
+            )
 
     async def _get_prompt(
         self, name: str = None, arguments: dict[str, str] = None