Your changelog,
one message away.
Connect the AI assistant you already use to ReleasePad and it can draft release notes, publish or schedule them, tidy up old posts and tell you what shipped, all from the conversation you are already in. Every post starts as a draft. Nothing goes public until you say so.
One URL, no API keys. Log in with your ReleasePad account and revoke any time.
What is the ReleasePad MCP server?
The ReleasePad MCP server is a connector, built on the Model Context Protocol, that lets AI assistants such as Claude, ChatGPT, Codex, Grok, Claude Code and Cursor operate your changelog directly. Once connected, the assistant can list your products, search and read posts, create and edit release notes, publish or schedule them, manage categories and read view analytics. It gives an agent the same access as the management API, without anyone writing code.
Write where the work happens
Paste a pull request, a ticket or three bullet points into the assistant you already use. It drafts the release note in your voice and saves it to ReleasePad. No tab switching, no blank page.
Ask your changelog questions
"What did we ship in August?" "Is the SSO post still a draft?" "Which release got the most views?" The assistant asks ReleasePad and answers from what is actually there.
Safe to hand to an agent
Every post starts as a draft. Destructive tools are flagged so clients like Claude confirm before deleting. Grant read-only access if you only want answers, and revoke any time from Connected Apps.
One URL, then log in
The server address is the same for every client. The first time a tool is used, a ReleasePad login page opens in your browser. You approve the access and the client keeps a token you can revoke at any time.
https://pro.releasepad.io/mcpAdd the server to your client
Paste the URL into Claude.ai, ChatGPT, Codex, Grok, Claude Code, Cursor or any client that supports remote MCP servers. Exact steps for each are below.
Log in and approve
A ReleasePad login page opens in your browser. Choose full access, or read-only if the assistant should only answer questions. That is the whole setup.
Ask for a release note
Describe what shipped. The assistant finds your product, picks a category, drafts the post and saves it. You review and say "publish."
Open Settings → Connectors, choose Add custom connector, name it ReleasePad and paste the URL. Click Connect and log in.
claude mcp add --transport http releasepad https://pro.releasepad.io/mcpThen run /mcp inside a session, pick releasepad and follow the login link.
Install it in one click from the ReleasePad listing in the Cursor directory, or add it by hand:
{
"mcpServers": {
"releasepad": {
"url": "https://pro.releasepad.io/mcp"
}
}
}Add it to ~/.cursor/mcp.json and restart. Cursor prompts you to log in the first time a tool is used.
Any client that supports remote MCP servers over Streamable HTTP with OAuth works the same way. The MCP Inspector (npx @modelcontextprotocol/inspector) is a good way to see every tool and try them by hand.
Things you can say on day one
Every one of these is a workflow the server supports today. Product names are examples; the assistant finds yours.
Here is the PR description for the CSV export change: [paste]. Write a release note for the Reports product, match the tone of our last three published posts, file it under Improvement and keep it as a draft.
The assistant reads your recent posts for voice, checks your categories, and saves a draft. You get a post to review, not text to copy.
After merging a user-facing change, draft a release note for it in ReleasePad (product: Reports) as a draft and tell me the post id. Never publish without asking.
Put this in your project's CLAUDE.md or Cursor rules. The agent that merged the change writes it up while the context is fresh.
What did we release in August? Group by category, one line per post. Then list any drafts older than two weeks.
Useful before a customer call, a board update or a "what's new" email. The answer comes from your changelog, not from memory.
Publish the "SSO for Google Workspace" draft. Schedule the "New billing page" draft for next Tuesday at 9:00 our time.
A future publish date schedules the post. ReleasePad publishes it on time to your changelog page and widget.
Read every draft in the Mobile app product. Make them friendlier, fix typos, keep each under 120 words. Leave them as drafts.
Each edit is a normal update, so nothing changes visibility. Ask the assistant to summarize what it changed afterwards.
Which post got the most views in the last 30 days, and where were the readers?
The assistant reads the same view analytics you see in the dashboard, so you know what deserves a longer post next time.
Fourteen tools, in plain terms
| Products | List your products and read one. Every other tool works inside a product. |
| Posts | List and search by status, title or category. Read one with its full Markdown body. Create, update, publish, schedule and delete. |
| Categories | List, create, rename, recolor and delete the categories posts are grouped under. |
| Analytics | List views of one post, or of the whole changelog page, with country and city. |
Every result is the same JSON the management API returns, so anything a script can do, a conversation can do.
Built to be handed to an agent
Common questions
Built and maintained by the ReleasePad team · Last updated
Any client that supports remote MCP servers over Streamable HTTP with OAuth login. That includes Claude.ai, Claude Desktop, Claude Code, ChatGPT, Codex, Grok, Cursor, and the MCP Inspector, plus most recent agent frameworks. The server address is the same for all of them: https://pro.releasepad.io/mcp. Cursor users can also install it in one click from the ReleasePad listing in the Cursor directory.
Only if you tell it to. Every post created through the MCP server is saved as a draft unless the assistant is explicitly asked to publish. Drafts are not visible on your changelog page or in your widget. If you want the assistant to never change anything, grant read-only access when you authorize it.
No. The first time a tool is used, the client opens a ReleasePad login page in your browser. You approve the access and the client keeps a token. You can see every connected client and revoke any of them under My Account, Connected Apps. Revoking ends that client's sessions immediately.
No, they work together. The GitHub integration watches your repository and drafts release notes from commits and pull requests automatically. The MCP server puts your changelog inside the assistant you already talk to, so you can also write from a ticket, a Slack thread or a few bullet points, ask what shipped last month, or reshape old posts.
Yes. Any MCP client library can connect to the same URL and run the same OAuth flow. Each tool has a typed input schema and a typed output schema, so agents get structured results rather than free text. If you would rather call HTTP endpoints directly, the management API exposes the same operations with an API key.
The MCP server is included in the paid plan, a flat $35 per month per product with every feature and unlimited team members. Free accounts can complete the connection but requests are refused until the account upgrades.
The release note you keep putting off is one message away.
Create an account, publish your first post, and connect your assistant when you want it to take over the writing. $35/mo per product, unlimited team members, every feature included.