Skip to content

Blog

How to keep Mintlify docs up to date when code changes

Mintlify already automates the structural half of keeping docs current: it builds API reference pages from your OpenAPI spec, checks links, validates builds, syncs with GitHub, and can draft pull requests when code merges. The half that slips is the prose and the examples whose facts live in other repositories, tickets, and threads.

EkLine and Mintlify logos side by side.
Published on
Read time
9 min

Mintlify already automates the structural half of keeping docs current: it builds API reference pages from your OpenAPI spec, checks links, validates builds, syncs with GitHub, and can draft pull requests when code merges. The half that slips is the prose and the examples whose facts live in other repositories, tickets, and threads. That half needs a review step, and it is where an EkLine pull request fits.

Follow one change through this post: an endpoint's auth header changes. The reference page generated from your spec updates. The quickstart and a guide still show the old header, and mint broken-links passes, because every link on those pages still resolves.

TL;DR

  • Mintlify regenerates API reference pages from your OpenAPI spec; a spec hosted at a URL needs a Trigger deployment call.
  • The Mintlify CLI checks links and build warnings, so an outdated header in valid MDX passes.
  • Mintlify automations draft pull requests after code merges, and Mintlify's own guidance still keeps review and sign-offs.
  • EkLine reads each code pull request as it opens and lands its docs pull request in the repository Mintlify deploys from.
Diagram: after one API change, the Mintlify API reference page generated from OpenAPI updates itself and link checks pass, while the quickstart and auth guide still show the old header; EkLine drafts the fix as a pull request and Mintlify deploys on merge

What Mintlify already keeps current

Generated API pages. Mintlify's OpenAPI setup lets you "Reference any number of OpenAPI specifications in the navigation element of your docs.json to create pages for your API endpoints." When the spec lives in your repository, a spec change is a Git push and the site redeploys. When you host the spec at a URL, Mintlify warns that "changes to the spec don't trigger a Git push, so your docs won't redeploy automatically," and tells you to call the Trigger deployment API endpoint in the same CI action that generates or updates the spec.

Link and build checks. In the Mintlify CLI, mint broken-links checks internal links; --check-anchors also validates anchor links against heading slugs, and --check-external also checks external URLs. mint validate "Validate[s] your documentation build in strict mode. Exits with an error if there are any warnings or errors."

Git sync. Mintlify "uses a GitHub App to automatically sync your documentation with your GitHub repository" and deploys when you push. In a monorepo, you can "specify a subdirectory if your docs are not at the repository root."

Drafted updates. Mintlify automations include a code change trigger that "Runs when a pull request merges in one or more connected source code repositories." The agent "reads your project content and any connected repositories, then follows the prompt to make updates." Automations require a Pro or Enterprise plan, "Each run that updates your docs costs 250 credits," and each automation can run up to 500 times per day. If the auth header change merges in a connected repository, this feature may already cover it, so check it before you add another tool.

Mintlify still keeps people in the loop. Its maintenance guide says "Update docs in the same pull request as the code change" and "Require a docs sign-off before closing any ticket that changes user-facing behavior," then sets monthly, quarterly, and annual reviews. A June 18, 2026 Mintlify article on stopping documentation drift adds that "pull request review is the safer setup for user-facing documentation."

Where the auth header change still slips through

Test your own repository. Open the last merged pull request that changed how your API authenticates or what a request must send. Did anything flag the quickstart or guide that shows the old request? If nothing did, the reason is in what each Mintlify check reads.

Native checkWhat it verifiesWhat it does not seeWhere EkLine fits
mint broken-linksInternal links, plus anchors and external URLs with flagsA working link on a page whose sample sends the old headerAutomatic PR review reads the API pull request and can start the docs fix
mint validateThe build in strict mode, failing on any warning or errorValid MDX that describes old behaviorDocs Reviewer comments on the docs change, with Mintlify component support
OpenAPI pagesReference pages generated from your specQuickstarts and guides written by hand in MDXDocs Agent edits those pages in your documentation repository

Two gaps sit outside every check: a migration note that lives in a ticket or a Slack thread, and hand-kept files. In a Mintlify-based repository, drewdrewthis reported on August 21, 2026 that llms-full.txt "is maintained by hand, which means it drifts silently and there is no signal when it does."

So when are native tools enough? When your API code and docs share a repository and every fact the guide needs is in the pull request, Mintlify's checks plus the same-PR rule from its maintenance guide cover the change. When the facts live in tickets, Slack, specs, or a separate code repository, the rule depends on memory; practitioners in an Ask HN thread on keeping docs current describe docs-first habits and homemade nightly scripts.

A Mintlify upkeep checklist for this week

You need write access to the docs repository and its CI. None of these steps needs EkLine.

  1. If your OpenAPI spec is hosted at a URL, call the Trigger deployment endpoint in the CI action that updates the spec.
  2. Run mint broken-links with --check-anchors and --check-external before you publish; renamed headings change their slugs.
  3. Add mint validate to CI so a warning fails the build before readers see it.
  4. When a header, parameter, or endpoint name changes, search the whole docs repository for the old string, including hand-maintained files such as llms-full.txt. The link checks pass on this failure.
  5. Adopt the two process rules from Mintlify's maintenance guide: docs in the same pull request as the code, and a docs sign-off before closing any ticket that changes user-facing behavior.
  6. If you run automations, keep their pull requests on review. Mintlify's drift article calls review the safer setup.

How EkLine plugs into a Mintlify repository

EkLine works in the repository Mintlify already deploys from, so your publishing path stays as it is.

QuestionOn Mintlify
Where your content livesThe repository Mintlify's GitHub App deploys from, at the root or in a monorepo subdirectory.
What EkLine watchesThe code repositories you select in the EkLine GitHub App, which lists them apart from your documentation repository. Docs Agent also reviews GitLab merge requests.
Where its pull request landsYour docs repository. Mintlify deploys on merge, as it does today.
Which CI check runsDocs Reviewer, which checks each docs change for style, consistency, links, and accuracy, with the mintlify framework value, set in its configuration or with --framework. The framework support page says "Mintlify currently has the most comprehensive component support."
What stays in Mintlifydocs.json, generated API pages, hosting, deploys, and any automations you run.

Walk the auth header change through Docs Agent. When the engineer opens the API pull request, or marks a draft ready, automatic PR review reads the diff, title, and description once and scores the documentation impact. A low score stays silent, a medium score leaves a suggestion comment, and a high score starts a docs pull request in your Mintlify repository for the pages that need the new header. You review and merge it, and Mintlify deploys. Automatic PR review reads the pull request alone and leaves linked tickets out. When the migration note sits in a Jira ticket or a Slack thread, open a Docs Agent session and combine the sources; if they disagree, "the agent flags the conflict in the chat so you can decide." A weekly Docs Drift Review sweeps what merged before you connected EkLine or scored below the threshold, and the Code Sample Review agent in Scheduled Agents "Verifies that code samples in your docs are accurate and runnable."

If your doubt is whether reviewers will trust drafts they did not write: over one quarter, April 7 to July 6, 2026, ContextQA's engineers "merged 97.5% of what the agent proposed, 79 updates, and set aside only two." ContextQA hosts its docs on GitBook, so read this as an EkLine result on another platform.

The limits: the assessment runs once per pull request, you cannot customize its confidence thresholds, and Docs Agent works with GitHub and GitLab. Docs Agent is available to all plans, and EkLine grants access on request.

FAQ

No. Mintlify keeps your docs.json, generated API reference, hosting, and deploys. EkLine adds a step before them: it reads code pull requests as they open, drafts docs pull requests into the repository Mintlify deploys from, and runs a Docs Reviewer check that handles Mintlify components. For a feature-by-feature view, read EkLine vs Mintlify.

Mintlify's own guidance keeps the review step. Its drift article says Mintlify "can be configured to automerge workflow pull requests, but pull request review is the safer setup for user-facing documentation," and its maintenance guide still asks for docs sign-offs and scheduled reviews. An automation drafts the change; a reviewer decides whether it is right before it goes live.

You can, and the Ask HN thread shows one version: cyanydeez described "a for loop that runs nightly on any files updated the prior day," with local AI revising the Markdown file next to each one. On Mintlify, when the quickstart lives in a separate docs repository, nothing sits beside the changed API file to revise. Someone also has to check each nightly rewrite, and rstagi named that cost in the same thread: "it's hard to fact-check everything all the time." EkLine reads the API pull request and, at high confidence, opens a docs pull request for the pages that need the change, so the check happens in review.


Bring one recent API pull request, such as the last auth change, and see what EkLine would have changed in your quickstart. Book an EkLine demo to walk through your own Mintlify setup.

Sources


Read more about

Blog

What Is LFM2.5-24B-A2B, and Can It Run Your Agent Locally?

·15 min read

LFM2.5-24B-A2B uses 2.45B parameters per token, but runtimes normally keep all 24B weights in memory. What hardware it needs, how its tool calls work, and a worksheet for deciding whether to adopt, test or defer it for a local agent.

Blog

How to keep a Pylon knowledge base accurate after every release

·11 min read

Pylon's signals catch what tickets show is missing. A release-time check catches what a release made wrong, before customers or the AI agent see the old answer. A playbook with three cadences and one approver per fix.

Blog

Claude Code for technical writers: what works and what breaks

·11 min read

A five-step Claude Code setup for documentation, from CLAUDE.md to a lint hook, plus an honest check of what the newer features fix and what still depends on one writer.

Blog

What Are GitHub Actions Lockfiles, and Do They Replace SHA Pinning?

·15 min read

GitHub Actions workflow lockfiles record the commit and source repository behind every action an onboarded workflow uses, transitive actions included. What they replace, what they leave to you, and how to pilot one release workflow.

Blog

AI Agent Pricing: What Should Count as a Billable Outcome?

·16 min read

Before you choose seats, usage, or outcome pricing for an AI agent, write down the event you will bill. A decision table, a billable-event record, seven edge cases, and a modeled cost per outcome.

Blog

What Is Poppy? When Should Your API Support Customer Agents?

·15 min read

Poppy, the Personal Agent Protocol from Sierra and Meta, is a draft 0.1 consent and session layer for customers' AI agents. What it adds over OAuth and MCP, and whether your API should adopt, experiment, or wait.