Very impressed with Rovo Dev, especially the agentic control capabilities (once, this session, forever etc). However there have been a few things have impacted my workflow.
1. Simpler addition of MCP servers.
2. Display git head in the status bar.
3. Plan Mode / Code Mode - Rovo Dev can sometimes be a little keen to dive into code. Plan mode would limit to only creating a plan.
4. Show plans / tasks as a simple markdown of work to be done / remaining.
5. /tools command so we can point Rovo Dev to the right tool when it is struggling.
6. When Rovo dev presents code / commands it needs to be in a markdown code window with a 'copy' option. Using the terminal select / copy picks up line breaks, box drawing characters etc, which can't be copied straight into code/terminal
3. /mode -plan is likely to consume less than /mode -code. This type of differentiation opens up the possibility of measuring / pricing the difference and informative for users.
Regarding MCP-Servers:
The json structure used is already the "standard" used by several AI tools on the market and by all MCP servers. Their documentation is a almost always covering which lines to add (use npm, uv, docker to connect). What I missed was a good tutorial, where the mcp.json file is located etc. - I still haven't found logging, for example if one MCP server crashes rovodev, like MS playwright-mcp does. I've reported a bug regarding this issue.
There is an interesting 3rd party project called "mcp-router" which offers visual configuration but acts a Server-in-the-middle in between the client (rovodev) and MCP serves. Have not audited their code so no guarantees whatsoever.