# ReleasePad Wants Your AI Coding Assistant To Write Your Release Notes Too > Source: > Published: 2026-10-05 19:36:23+00:00 *The feature ships, the pull request merges, and the one task that would tell anyone about it sits on a to-do list until it stops mattering.* That is the problem ReleasePad set out to close with the launch of a Model Context Protocol (MCP) server, announced this week, that lets AI coding assistants such as Claude, Codex and Cursor draft a product's release notes directly from the development workflow. The server is live at pro.releasepad.io/mcp and works with any assistant that supports the protocol. ## The task nobody owns ReleasePad's own framing of the problem is blunt: most product teams ship every week, but very few announce every week. The gap is not a shortage of tools for writing announcements, it is that writing them is nobody's job. The engineer who built the feature has already moved to the next one. The product manager is in planning. The marketer does not know the feature shipped at all. So the changelog goes quiet, users start to assume the product has stalled, and support ends up fielding questions that a two-line note would have answered. That silence compounds in a specific way. A missed release note is not a one-time cost, it is a recurring one: every week the changelog stays quiet is another week a prospective customer checking activity before a purchase decision sees nothing, another week an existing user assumes a bug they reported was never addressed, another week support answers the same "is this still being worked on" question that a public note would have closed off. None of that shows up as an obvious failure at the time. It shows up later, diffused across support tickets, churn conversations and sales calls, in a form that is hard to trace back to the missing note that caused it. Some teams have already started working around this the manual way, pasting commit messages into an AI assistant and asking for a summary after the fact. That approach works, but it depends on someone remembering to do it, on a separate context switch away from the tool where the work actually happened, and on that person having enough context on the change to write it up accurately days or weeks after merging. ReleasePad's bet is that the more useful version of that workflow sits inside the tools engineers already have open, not bolted on as a separate step once the code is done. [A Solo Coder Spent $10,000 in AI Tokens to Build a GTA Clone of Taipei](https://startupfortune.com/a-solo-coder-spent-10000-in-ai-tokens-to-build-a-gta-clone-of-taipei/) The game's creator, who goes by aicodewithme, built it using over 30 AI models including DeepSeek and Kimi, routed through a personal platform called AICodeWith, working five to six hours a day for roughly a month. - [how to build GTA clone using AI models](https://startupfortune.com/a-solo-coder-spent-10000-in-ai-tokens-to-build-a-gta-clone-of-taipei/) - [solo developer AI tokens game development costs](https://startupfortune.com/a-solo-coder-spent-10000-in-ai-tokens-to-build-a-gta-clone-of-taipei/) ## How it fits into the workflow The MCP server approach means the release-note generation is not a standalone app a team has to remember to visit. Claude, Codex and Cursor are already where the code gets written, reviewed and merged, so plugging a [Changelog tool](https://releasepad.io) in at that layer turns an announcement from a task someone has to remember into something the AI stack can surface as part of finishing the work. In the ReleasePad model, the assistant drafts the note and a human still approves it before anything goes out, which keeps the accountability where teams want it while removing the blank-page problem that usually kills the habit of announcing. That approval step matters for adoption as much as it does for accuracy. Teams that have tried fully automated changelog generation before tend to distrust it, because a wrong or overly technical note published without review can do more damage to user trust than no note at all. By keeping a human in the loop at the point of publishing rather than at the point of drafting, ReleasePad is trying to remove the part of the task that is tedious, the first pass at turning code changes into plain language, while leaving the part that requires judgment, deciding what is worth saying and how to say it, in human hands. Once approved, the note publishes out to wherever the team's changelog lives, which is the second half of the gap ReleasePad is targeting: it is not just that nobody writes the note, it is that even a written note often stays buried in a doc nobody checks. Turning the note into something that goes out automatically once it is signed off closes both ends of the problem at once. ## Why this matters beyond one product The pattern ReleasePad is chasing is a familiar one in how AI coding assistants are being adopted more broadly: the tools are moving from being asked questions to being embedded in the actual mechanics of shipping software, one narrow, well-defined task at a time. Release notes are a good candidate for that kind of embedding precisely because the task is low-stakes to get slightly wrong, repetitive enough to be tedious, and blocked less by difficulty than by nobody claiming ownership of it. A team that ships weekly but rarely announces is not failing at communication out of neglect, it is failing because the step never had a natural owner in the first place. That ownership gap tends to widen as a team grows. In a two-person startup, whoever merges the change can also dash off a note about it without much friction. Once a company has separate engineering, product and marketing functions, the information about what shipped and the responsibility for announcing it live in different people's heads, and the step that used to happen informally has nowhere obvious to land. Embedding the drafting step directly in the assistant that already has visibility into the commits and the pull request sidesteps that handoff problem rather than trying to fix it with more process. For product and engineering teams that already treat their release history as part of the product experience, an MCP server sitting inside Claude, Codex or Cursor removes the excuse that writing the note takes too long or breaks focus. The commit messages and the pull request already contain most of what a release note needs to say. What ReleasePad is offering is the plumbing to turn that raw material into something a customer would actually want to read, without asking anyone to open a separate tool to do it. ReleasePad's site is at releasepad.io, where teams can find the MCP server details and get it connected to the assistant they already use. [A wave of narrow AI inference engines is beating vLLM and llama.cpp at their own game](https://startupfortune.com/a-wave-of-narrow-ai-inference-engines-is-beating-vllm-and-llamacpp-at-their-own-game/) Engines like Strata, ninfer, and Splash beat llama.cpp by up to 4x on Qwen3.8-Flash-Next, but each works for exactly one model on one GPU family, a tradeoff buyers often miss. - [narrow AI inference engines outperforming vLLM](https://startupfortune.com/a-wave-of-narrow-ai-inference-engines-is-beating-vllm-and-llamacpp-at-their-own-game/) - [specialized GPU runtime for local AI models](https://startupfortune.com/a-wave-of-narrow-ai-inference-engines-is-beating-vllm-and-llamacpp-at-their-own-game/) ## Join the discussion [Open in the community →](https://startupfortune.com/community/) Almost there. Sign in and your reply posts straight away.