One connection. A real first result.
Add the server, sign in and approve, then verify your workspace connection. No key to copy for an OAuth-capable agent.
Choose the OpenAI Agents SDK Montage execution boundary
Before adapting connection code, decide whether the application or Responses API should execute remote tools.
Read guideSeparate OpenAI service credentials from Montage authorization
An application can authenticate successfully to OpenAI while having no approved Montage account access.
Read guideReview OpenAI Agents SDK local MCP connection lifetime
A locally controlled SDK connection must stay valid while its discovered tools are used.
Read guideReview Montage tool approval before an OpenAI Agents SDK run
A remote tool being available to an SDK agent does not mean its effects were approved.
Read guideReview Pydantic AI MCPToolset generation before adapting Montage
Pydantic AI examples from different API generations should not be mixed into one application.
Read guideIsolate Pydantic AI Montage authorization by user session
A Pydantic AI toolset can wrap an authorized FastMCP client, but it must not become a vehicle for sharing one user’s workspace access with others.
Read guideKeep Pydantic AI Montage tools inside their async lifetime
A discovered tool can depend on a client that closes before the Agent ever uses it.
Read guideGate Pydantic AI Agent tool use on a direct Montage read
Attaching a toolset is not the same as proving an authorized account can execute it.
Read guideReview LlamaIndex tool filtering for a Montage status diagnostic
A LlamaIndex application can discover many remote tools when it needs only one preliminary read.
Read guideReview a LlamaIndex custom ClientSession authorization boundary
LlamaIndex documents custom ClientSession integration, but it is an extension point rather than a shortcut around protocol initialization or account authorization.
Read guideMove LlamaIndex Montage tool use into the HTTP client lifetime
LlamaIndex discovery helpers can make remote tools look like ordinary reusable objects.
Read guideProduce a LlamaIndex status receipt without a retrieval or agent run
A notebook can display a discovered tool list and still have no successful Montage workspace call.
Read guideSelect Semantic Kernel’s Python Streamable HTTP plugin for Montage
Semantic Kernel’s main MCP example may show SSE, which is not the hosted transport used here.
Read guideReview Semantic Kernel plugin authorization independently of model access
Semantic Kernel may have valid model-service configuration while its Montage plugin lacks authorized account access.
Read guideKeep Semantic Kernel Montage functions within the plugin lifetime
Discovered kernel functions can outlive the remote plugin that supplied them if application cleanup is misplaced.
Read guideReview Montage function exposure before a Semantic Kernel model run
Adding MCP functions to a kernel can broaden what a model may propose.
Read guideSelect an installed MCP Inspector interface without downloading tools
Inspector examples can invoke package runners that fetch dependencies automatically.
Read guideSeparate Inspector proxy access from Montage authorization
Inspector may have a protected local proxy as well as a remote Montage account challenge.
Read guideTrace Montage initialization discovery and execution in MCP Inspector
A failed MCP workflow may break during initialization, discovery or tool execution.
Read guidePrepare a secret-safe Montage status handoff from MCP Inspector
Inspector is useful for handing protocol evidence to another engineer, but raw transcripts may contain local proxy or account secrets.
Read guideDistinguish FastMCP client construction from an active Montage session
A FastMCP Client object can be constructed before the application ever establishes a usable remote session.
Read guideReview FastMCP OAuth storage before serving multiple Montage users
FastMCP can supply direct remote access without a model, but private OAuth state must remain tied to its owner.
Read guideHandle a FastMCP Montage call failure without leaking its client lifetime
A failed direct call should leave an explicit error and a properly released client, not a reusable-looking success object.
Read guideInspect the FastMCP Montage tool contract before direct execution
A direct FastMCP diagnostic can stay narrower than an agent workflow while still checking the actual remote contract.
Read guide