Hello, I was using some agents at the "Composio for you" platform, connecting directly my claude and agents directly to the Composio MCP server. After the incident, I got the same issue as everyone, server returning 500 errors etc etc. Then I found that "Composio Platform", api keys were working so I just told Claude code to setup a local MCP server using the composio platform api key. Then I added the toolkits that I needed and reconnected my agents, now the agents runs even a bit faster, since the MCP is hosted on my localmachine. Just sharing to people that may be struggling to restore their agents Good luck everyone

Hi @Eduardo Nuri, thanks fot the workaround! Going local-MCP with a Platform API key is a solid route while we've got For You temporarily down to finish restoration post our [security incident](https://composio.dev/blog/composio-may-2026-security-incident). Mileage might vary a bit depending on the toolkits and setup.

Hey @Eduardo Nuri would you be so kind as to share the mcp code or just give a bit more detail on what you did so we can replicate? Did you just copy the consumer MCP and swap keys? Or was there something else involved?

I just told claude code the whole situation in plan mode something like; we've our App that used the "Composio for you" app we've a working api key at "Composio platform" other app at composio, research the web about both products, see the differences, see if you wan setup a composio platform MCP server to connect my apps, test it. Then when the AI finished coding it and I put the api key and I connected my apps, it asked me if I wanted to switch MCP urls from the composio for you MCP cloud server to my local MCP server, and I found that it works even a bit faster.