RTM MCP - I can talk to RTM in plain English
Hi,
I have built an MCP server for RTM (available as open source to all: https://github.com/ljadach/rtm-mcp) and integrated it with my personal AI assistant on Claude Code. Now together with the Assistant we do the daily planning and day closing and it has full access to my ToDo list from RTM, so I can, in conversational style, discuss my priorities ("what are the top tasks for today"), ask it to make adjustments ("change the due date of all tasks that ....) and create the task automatically resulting from my discussion with AI assistant or from meeting notes processed by AI.
In the evenings, thanks to the MCP integration the Assistant can identify patterns ("you have been postponing this task for 3 days now"), praise me ("you have completed x tasks and I am proud of you for completing x"). It can also maintain the activity log and check if I am progressing against my long term goals and objectives.
I think this is only the beginning. The possibilities of integrating AI with RTM are endless!
I have built an MCP server for RTM (available as open source to all: https://github.com/ljadach/rtm-mcp) and integrated it with my personal AI assistant on Claude Code. Now together with the Assistant we do the daily planning and day closing and it has full access to my ToDo list from RTM, so I can, in conversational style, discuss my priorities ("what are the top tasks for today"), ask it to make adjustments ("change the due date of all tasks that ....) and create the task automatically resulting from my discussion with AI assistant or from meeting notes processed by AI.
In the evenings, thanks to the MCP integration the Assistant can identify patterns ("you have been postponing this task for 3 days now"), praise me ("you have completed x tasks and I am proud of you for completing x"). It can also maintain the activity log and check if I am progressing against my long term goals and objectives.
I think this is only the beginning. The possibilities of integrating AI with RTM are endless!
This is awesome — love seeing RTM get some AI love! Big props for publishing the first open-source RTM MCP and sharing your daily planning + evening reflection workflow. That pattern detection ("you've been postponing this for 3 days") is exactly the kind of thing that makes AI assistants actually useful rather than just cool demos.
Your post actually inspired me to build my own take: milk-mcp (https://github.com/donkitchen/milk-mcp). Same DNA — TypeScript, RTM, Claude Code — but I took it in a different direction. Where your setup shines for personal productivity and conversational task management, I focused specifically on the developer workflow problem: Claude Code sessions are ephemeral, and losing context between sessions is brutal. milk-mcp turns RTM into a persistent memory layer so Claude picks up exactly where it left off.
It creates 5 structured lists per project (TODOs, Backlog, Bugs, Decisions, and a Context/handoff list) and has dedicated tools for session_start and session_end so the handoff is explicit and reliable. The Decisions list is my personal favorite — architectural choices get logged with rationale, so Claude never asks "why did we do it this way?" again.
Two different tools solving related problems — would love to see the community build on both! 🚀
Your post actually inspired me to build my own take: milk-mcp (https://github.com/donkitchen/milk-mcp). Same DNA — TypeScript, RTM, Claude Code — but I took it in a different direction. Where your setup shines for personal productivity and conversational task management, I focused specifically on the developer workflow problem: Claude Code sessions are ephemeral, and losing context between sessions is brutal. milk-mcp turns RTM into a persistent memory layer so Claude picks up exactly where it left off.
It creates 5 structured lists per project (TODOs, Backlog, Bugs, Decisions, and a Context/handoff list) and has dedicated tools for session_start and session_end so the handoff is explicit and reliable. The Decisions list is my personal favorite — architectural choices get logged with rationale, so Claude never asks "why did we do it this way?" again.
Two different tools solving related problems — would love to see the community build on both! 🚀
Log in
to post a reply.