Write end-of-sprint changelogs from completed tickets
i'm a PM and the task i hated most was the end-of-sprint changelog. Task: i'm a PM and the task i hated most was the end-of-sprint changelog. every two weeks i'd spend an hour sifting through completed linear tickets, deciding what's worth mentioning, writing it up in chatgpt, publishing it, then figuring out if it warrants an email to users or an in-app announcement. tedious, repetitive, and always the thing i'd procrastinate on. set up a cowork task to do the whole thing. runs every two weeks automatically. claude connects to linear via MCP, pulls completed issues, figures out which ones are user-facing, writes the changelog copy using the actual ticket context, and publishes it through another MCP connection. if the update is big enough it triggers an email and in-app notification too. smaller stuff just goes to the changelog page quietly. the part that surprised me: the copy is genuinely better than what i was writing manually. claude pulls details from ticket descriptions and comments that i would've skipped because i was rushing to get it done. 90% of the time i just review and ship it. only thing i still do by hand is the header image. 2 minutes with a screenshot beautifier. i think cowork is undersold as a scheduling tool honestly. most of the use cases i see are one-off tasks but the real power is the recurring stuff. the boring work that eats an hour every week or every sprint that you never get around to automating because writing a script feels like overkill. cowork just lets you describe what you want in plain english and schedule it. what recurring tasks are you running with cowork? curious what else people have automated beyond coding workflows.; tools: Claude Cowork. Human role: The original post does not say how a person stayed involved.; limitations: Source reliability is low as the original URL returned a 403 error, preventing verification. The workflow assumes access to Linear and MCP integrations. Privacy and compliance concerns were raised by readers but not addressed in the text. Human approval is required before publishing..
Tools
Published on Agent Work Atlas:
Overview
i'm a PM and the task i hated most was the end-of-sprint changelog.
Workflow
i'm a PM and the task i hated most was the end-of-sprint changelog. every two weeks i'd spend an hour sifting through completed linear tickets, deciding what's worth mentioning, writing it up in chatgpt, publishing it, then figuring out if it warrants an email to users or an in-app announcement. tedious, repetitive, and always the thing i'd procrastinate on. set up a cowork task to do the whole thing. runs every two weeks automatically. claude connects to linear via MCP, pulls completed issues, figures out which ones are user-facing, writes the changelog copy using the actual ticket context, and publishes it through another MCP connection. if the update is big enough it triggers an email and in-app notification too. smaller stuff just goes to the changelog page quietly. the part that surprised me: the copy is genuinely better than what i was writing manually. claude pulls details from ticket descriptions and comments that i would've skipped because i was rushing to get it done. 90% of the time i just review and ship it. only thing i still do by hand is the header image. 2 minutes with a screenshot beautifier. i think cowork is undersold as a scheduling tool honestly. most of the use cases i see are one-off tasks but the real power is the recurring stuff. the boring work that eats an hour every week or every sprint that you never get around to automating because writing a script feels like overkill. cowork just lets you describe what you want in plain english and schedule it. what recurring tasks are you running with cowork? curious what else people have automated beyond coding workflows.
Reported result
The author reports that the automated copy is often better than manual drafts because it includes details from comments. They state they review and ship the output 90% of the time. No independent metrics or failure rates are provided in the source.
Human role
The original post does not say how a person stayed involved.
Limitations
Source reliability is low as the original URL returned a 403 error, preventing verification. The workflow assumes access to Linear and MCP integrations. Privacy and compliance concerns were raised by readers but not addressed in the text. Human approval is required before publishing.
Try it
Act as a Product Manager automating sprint changelogs. Use Claude Cowork with Linear MCP. Input: [Linear Project ID], [Sprint Dates], [Changelog URL]. Steps: 1. Pull completed issues. 2. Filter for user-facing changes. 3. Draft changelog copy using ticket descriptions and comments. 4. Identify major updates for email/announcement. 5. Generate a draft for review. Do not publish or send emails without my explicit approval. First, run a dry test on [Previous Sprint ID] and show me the draft and proposed notifications. Wait for my confirmation before proceeding to the current sprint. Ensure no sensitive internal data is included in the public copy.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.