Hey everyone! I'm exploring a way to centralize agent tools in an enterprise environment. My goal is to have agent logic and execution running in the cloud while maintaining the same developer interface and workflow like Kiro UI, Kiro-cli, and Claude Code. Essentially, I want developers to interact from their local machines using the native interfaces, but I want everything—prompts, tools, versions—to be managed from a central location and shared among all developers. I'm looking for solutions that don't require creating a custom UI or having agents run locally for each developer. Are there current agent platforms that support this type of setup? Any examples of tools or architectures that could help? Thanks!
3 Answers
We actually do something similar using Codex. You just deploy the model on Azure and set up the client to connect to that endpoint. At least for Codex, the documentation on how to configure the client is pretty solid.
Are you considering deploying the Claude model via Azure Foundry? I imagine AWS Bedrock would work similarly. Just earlier today, I was looking into the performance of Anthropic models across different platforms.
Actually, I’m focusing specifically on Claude-Code, not just any Claude model. I want to set it up in the cloud with a certain configuration like a system prompt, so every developer can access it.
Have you thought about using Claude's capabilities to help with this? I read somewhere that it can be set up in a way to meet your needs.
I see what you mean, but I think OP is looking for a slightly different approach. In your setup, the MCP server and tool invocations still come from the local device. OP seems to want all requests to go through a central system that distributes them to the necessary MCP servers or tools.