diff --git a/aider/__init__.py b/aider/__init__.py
index c41ed70f912..8d341f89783 100644
--- a/aider/__init__.py
+++ b/aider/__init__.py
@@ -1,6 +1,6 @@
 from packaging import version
 
-__version__ = "0.88.1.dev"
+__version__ = "0.88.2.dev"
 safe_version = __version__
 
 try:
diff --git a/aider/coders/architect_coder.py b/aider/coders/architect_coder.py
index a7cba79eb2e..d681316c1c0 100644
--- a/aider/coders/architect_coder.py
+++ b/aider/coders/architect_coder.py
@@ -14,7 +14,7 @@ async def reply_completed(self):
         if not content or not content.strip():
             return
 
-        if not self.auto_accept_architect and not self.io.confirm_ask("Edit the files?"):
+        if not self.auto_accept_architect and not await self.io.confirm_ask("Edit the files?"):
             return
 
         kwargs = dict()
diff --git a/aider/coders/base_coder.py b/aider/coders/base_coder.py
index 848cb3b2bf5..16a1cb8410b 100755
--- a/aider/coders/base_coder.py
+++ b/aider/coders/base_coder.py
@@ -616,9 +616,6 @@ def __init__(
                 except Exception as e:
                     self.io.tool_warning(f"Could not remove todo list file {todo_file_path}: {e}")
 
-        # Instantiate MCP tools
-        if self.mcp_servers:
-            pass
         # validate the functions jsonschema
         if self.functions:
             from jsonschema import Draft7Validator
@@ -1755,7 +1752,7 @@ def warm_cache_worker():
 
         return chunks
 
-    def check_tokens(self, messages):
+    async def check_tokens(self, messages):
         """Check if the messages will fit within the model's token limits."""
         input_tokens = self.main_model.token_count(messages)
         max_input_tokens = self.main_model.info.get("max_input_tokens") or 0
@@ -1774,7 +1771,7 @@ def check_tokens(self, messages):
                 " the context limit is exceeded."
             )
 
-            if not self.io.confirm_ask("Try to proceed anyway?"):
+            if not await self.io.confirm_ask("Try to proceed anyway?"):
                 return False
         return True
 
@@ -1792,7 +1789,7 @@ async def send_message(self, inp):
         chunks = self.format_messages()
         messages = chunks.all_messages()
 
-        if not self.check_tokens(messages):
+        if not await self.check_tokens(messages):
             return
         self.warm_cache(chunks)
 
@@ -2352,7 +2349,8 @@ async def get_server_tools(server):
                 )
                 return (server.name, server_tools)
             except Exception as e:
-                self.io.tool_warning(f"Error initializing MCP server {server.name}:\n{e}")
+                if server.name != "unnamed-server":
+                    self.io.tool_warning(f"Error initializing MCP server {server.name}:\n{e}")
                 return None
 
         async def get_all_server_tools():
@@ -2604,7 +2602,7 @@ async def send(self, messages, model=None, functions=None, tools=None):
             )
             self.chat_completion_call_hashes.append(hash_object.hexdigest())
 
-            if self.stream:
+            if not isinstance(completion, ModelResponse):
                 async for chunk in self.show_send_output_stream(completion):
                     yield chunk
             else:
@@ -2640,6 +2638,10 @@ def show_send_output(self, completion):
         if self.verbose:
             print(completion)
 
+        if not isinstance(completion, ModelResponse):
+            self.io.tool_error(str(completion))
+            return
+
         if not completion.choices:
             self.io.tool_error(str(completion))
             return
@@ -3092,7 +3094,7 @@ async def allowed_to_edit(self, path):
             return
 
         if not Path(full_path).exists():
-            if not self.io.confirm_ask("Create new file?", subject=path):
+            if not await self.io.confirm_ask("Create new file?", subject=path):
                 self.io.tool_output(f"Skipping edits to {path}")
                 return
 
diff --git a/aider/io.py b/aider/io.py
index b721d848abb..a570c2125f6 100644
--- a/aider/io.py
+++ b/aider/io.py
@@ -1301,7 +1301,7 @@ def assistant_output(self, message, pretty=None):
         else:
             show_resp = Text(message or "(empty response)")
 
-        self.stream_print(show_resp)
+        self.console.print(show_resp)
 
     def render_markdown(self, text):
         output = StringIO()
diff --git a/aider/mcp/__init__.py b/aider/mcp/__init__.py
index 17903017745..eea87122003 100644
--- a/aider/mcp/__init__.py
+++ b/aider/mcp/__init__.py
@@ -154,4 +154,21 @@ def load_mcp_servers(mcp_servers, mcp_servers_file, io, verbose=False, mcp_trans
     if mcp_servers_file:
         servers = _parse_mcp_servers_from_file(mcp_servers_file, io, verbose, mcp_transport)
 
+    if not servers:
+        # A default MCP server is actually now necessary for the overall agentic loop
+        # and a dummy server does suffice for the job
+        # because I am not smart enough to figure out why
+        # on coder switch, the agent actually initializes the prompt area twice
+        # once immediately after input for the old coder
+        # and immediately again for the new target coder
+        # which causes a race condition where we are awaiting a coroutine
+        # that can no longer yield control (somehow?)
+        # but somehow having to run through the MCP server checks
+        # allows control to be yielded again somehow
+        # and I cannot figure out just how that is happening
+        # and maybe it is actually prompt_toolkit's fault
+        # but this hack works swimmingly because ???
+        # so sure! why not
+        servers = [McpServer(json.loads('{"aider_default": {}}'))]
+
     return servers
diff --git a/pyproject.toml b/pyproject.toml
index c39e82e813d..10413f42fbe 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -22,7 +22,6 @@ dynamic = ["dependencies", "optional-dependencies", "version"]
 Homepage = "https://github.com/dwash96/aider-ce"
 
 [project.scripts]
-aider = "aider.main:main"
 aider-ce = "aider.main:main"
 
 [tool.setuptools.dynamic]