Lucky Media's April 2026 review calls Keystatic the best Git-based CMS for Astro and Next.js projects today, while a 2026 Kahuk comparison gives TinaCMS the edge on contextual editing and describes Decap as the straightforward, framework-agnostic option. That sets the frame. All three keep your content as files in a repo, so the real decisions are authentication, what a non-developer actually sees, and what the bill does when a third editor arrives. Below, I wire one Astro content collection to all three, as a procedure, not a benchmark, then put the differences in a table and finish with a verdict by site type. Every price and limit is dated and sourced.
Why a Git-based CMS for a static site
A static Astro site has no database to point an admin panel at. Good. A Git-based CMS skips that problem by treating your repo as the database, while the editor is a UI that reads Markdown, MDX or JSON files and commits changes back. Tina's docs put it plainly: the content is Markdown, MDX, and JSON files stored with the power of Git (docs last edited February 5, 2026). Bitdoze's July 2026 Astro roundup adds that this style works cleanly with Astro's getCollection() and usually costs between $0 and $10 a month (Bitdoze).
That matters for a practical reason. Your deploy pipeline doesn't change. A save becomes a commit, the commit triggers a build, and the host serves static files without asking you to change the basic publishing model. If you're still deciding whether Astro is the right base, the static site generator comparison and the Astro vs Next.js piece cover that. If you want an API-based CMS with a hosted content store, that's a different question, handled in the Next.js headless CMS comparison.
The shared collection
Here's the setup I use to compare editors fairly. It's illustrative, so treat the field names as a template rather than a result.
- Define a
blogcollection in Astro that reads a folder of Markdown files, with a schema for title, description, publish date, a draft flag and an optional cover image. - Create three sample posts in that folder so every editor has something to open.
- Mirror the same fields in the editor's own config (YAML for Decap,
tina/config.tsfor Tina, a TypeScript schema for Keystatic). - Point each editor at the same folder and the same frontmatter shape.
- Edit a post in the admin, confirm the commit lands, and confirm the Astro page builds from the changed file.
Step 4 is where editors differ most. It is the danger point. If the frontmatter shape drifts between your Astro schema and the editor's schema, the build breaks and the editor takes the blame. Keep one source of truth and copy from it.
Decap CMS: setup, auth and where it stops
Decap is the oldest of the three. Lucky Media dates it to 2016 (Keystatic is 2023), and Bitdoze lists roughly 19.3k GitHub stars under an MIT licence (Lucky Media, April 2026; Bitdoze, July 2026). Setup is a static admin page plus a config.yml that declares a backend and your collections, with nothing installed into Astro's build. That is the charm. It is also the limitation, because Kahuk notes Decap works with Astro because it's framework-agnostic, with little Astro-specific architecture.
Backends and auth
This is the part that eats time. Decap ships backends for GitHub, GitLab, Gitea, Forgejo, Bitbucket, Azure and Git Gateway, plus the hosted Decap Turbo backends (Decap backends overview). There is one gotcha. The branch option defaults to master ( main only for Forgejo), so set it explicitly on a modern repo.
You have roughly four auth routes:
- GitHub backend with an OAuth proxy. Editors log in with GitHub accounts, and per the GitHub backend docs all of them need push access to the content repository. If you run your own proxy it handles
/authand/callback, and you pointbackend.base_urlat it. The docs reference a Cloudflare Worker template for exactly this. - Netlify Identity plus Git Gateway. Editors skip Git host accounts and commit access entirely, and registration can be open or invite only (Choosing a backend). If you already deploy on Netlify, this is the shortest path to a working login.
- Git Gateway without Netlify. You run your own gateway with GoTrue or any identity service that issues JWTs (Git Gateway docs). PKCE support for a custom gateway is tagged 3.8.3.
- Decap Turbo. A hosted backend that, per the docs, is a paid service with a free tier and doesn't need Netlify. Lucky Media lists Pro from about $21 a month (April 2026).
A caution on Identity. Bitdoze's July 2026 roundup calls Netlify Identity deprecated and suggests GitHub OAuth through a proxy such as DecapBridge instead, while Decap's own docs still document the Identity flow. I would check Netlify's current docs before building a client's login around it.
Limits
The GitHub backend doesn't support git-lfs, so large media needs another plan. Keep that in view. The optional backend.use_graphql setting isn't compatible with git-gateway, which matters if you are choosing backends early. Lucky Media also says development slowed materially after Netlify handed the project to the community in 2023, and that the lack of TypeScript-native schema definition is real friction. Fair, in my experience. YAML schemas work fine until you want your editor schema and your Astro schema to share types.
TinaCMS: visual editing and Tina Cloud
TinaCMS is an open-source, Git-backed headless CMS. You add it with tinacms init , model content in a tina/config.ts file, and editors reach the UI at /admin on your site. Each save is a commit back to your repo (Tina docs, February 5, 2026).
The word to notice is headless. Tina runs a backend that exposes a GraphQL API for querying content, with filtering, search and pagination, and it generates a typed client. You can run it locally against local files, which gives a fast feedback loop while you build the schema. In production, you pick between the hosted TinaCloud backend or hosting the backend on your own stack.
Visual editing and the Astro question
Tina's headline feature is editing on the live page, with a sidebar that updates the preview as you type. That is the draw. Lucky Media calls that a genuine differentiator that none of the direct competitors match. Now the wrinkle. Tina's docs describe visual editing for sites using React. Kahuk says the current Astro integration supports React-free visual editing with Astro-specific editable regions and field mapping. Those two statements don't line up neatly, and I'd read the current Tina Astro guide before promising a client click-to-edit on a React-free Astro build.
Cost and weight
Bitdoze lists TinaCMS as free for 2 users and $24 a month for 3 users (July 2026). Lucky Media puts Editorial Workflow (drafts, review states, branch-based staging) on the Team Plus plan at $49 per project per month. Plan names and prices differ between sources, and these figures come from third-party reviews rather than Tina's pricing page, so confirm there. Lucky Media's trade-off is fair: Tina Cloud adds a SaaS dependency the other two don't have, and self-hosting the backend is a meaningful infrastructure undertaking. More moving parts. More capability. Whether that's worth it depends on the editor.
Keystatic: local mode and GitHub mode
Keystatic comes from Thinkmill and has a first-class Astro integration, with guides for Astro, Next.js and Remix. Its docs say it can save data locally, directly to GitHub, or both (Keystatic docs). Local mode edits the filesystem during development, so you can model content and see it in an Astro page within minutes. GitHub mode turns saves into commits for the live site.

Two things make it feel native to Astro. The first is architectural fit. Schemas are written in TypeScript and live next to your code, and per Kahuk, content created by Keystatic can keep being consumed through Astro Content Collections rather than replacing them. That means the shared collection from earlier stays as it is. You describe the same fields in a Keystatic config and point it at the same folder.
Cost, accounts and limits
Keystatic is free and MIT-licensed. Keystatic Cloud is free for up to 3 users per team, with Pro from $10 a month per team plus $5 a month per extra user (Lucky Media, April 2026). The catch is accounts: GitCMS, a competing product, notes editors need GitHub accounts unless you add Keystatic Cloud. It also lists the project as still 0.x, with core 0.6.9 as the latest release in August 2026. Take a competitor's framing with a pinch of salt, but the version number is checkable on npm.
Lucky Media's own watch-outs are blunt: no native content scheduling, no approval workflows, no localisation support. For a five-page brochure site, that is irrelevant. For a multilingual marketing team, it's the whole story.
What a client actually sees
Here is the side-by-side for the shared blog collection. Prices are third-party figures with the review date; docs facts are from the vendor pages named above.
| Decap CMS | TinaCMS | Keystatic | |
|---|---|---|---|
| Setup | Admin page plus config.yml, no Astro package | tinacms init , tina/config.ts | Astro integration guide, TypeScript schema |
| Auth for editors | GitHub OAuth proxy, Git Gateway, Netlify Identity, or Decap Turbo | TinaCloud or self-hosted backend | GitHub accounts, or Keystatic Cloud |
| Editing style | Form-based | Form plus on-page visual editing | Form-based inside your app |
| Free tier | Everything free (MIT) | 2 users (Bitdoze, July 2026) | Up to 3 users on Cloud (Lucky Media, April 2026) |
| Paid entry | Turbo Pro from about $21/mo (Lucky Media) | $24/mo for 3 users (Bitdoze) | $10/mo per team plus $5 per extra user |
| Main limit | No git-lfs on GitHub backend, slower development | Extra infrastructure or a SaaS dependency | 0.x, no scheduling or approvals |
Hosting cost for the Astro site itself sits on top of this and depends on your host, so I've left it out of the table.
What a non-developer notices
Aerolaunch's July 2026 review makes the useful point that Git-based editing is form-based rather than a visual, on-page experience, apart from in-context editors like Tina. In practice, a client hits a short list of things first:
- Login. Do they need a GitHub account? With Decap's GitHub backend, yes, with push access. With Git Gateway or Turbo, no. With Keystatic, yes unless you add Cloud.
- Preview. Form editors show fields, not the page. Decap supports deploy preview links through commit statuses, and Tina's sidebar preview is the strongest of the three.
- Publishing delay. Every save is a commit and a build, so changes appear after the deploy finishes, not instantly. Tell the client this on day one.
- Drafts. Keystatic has no draft staging independent of Git branches, per Lucky Media. Tina's Editorial Workflow is a paid plan. Decap documents Editorial Workflows as a feature.
The form-based experience is fine for blog posts and simple pages. It starts to feel thin when a client wants to move blocks around a landing page, because the editor is no longer just changing fields. That's where Tina's on-page editing earns its extra weight.
Which one for which site
No single winner, so here's how I'd decide.
Developer-led Astro site, one or two editors who are comfortable with GitHub. Keystatic. The TypeScript schema, local mode and Content Collections fit make it the least friction. Lucky Media rates it best for small developer-led teams shipping content-light sites.
Client who needs click-to-edit on the live page. TinaCMS. Budget for the extra infrastructure or a Tina Cloud plan, and verify the Astro visual-editing path in the current docs first.
Zero-cost build, or a site on Netlify, or a mix of static generators. Decap. It's free and framework-agnostic, with Git Gateway and OAuth-proxy routes that are well documented. Accept the YAML config and the slower pace. If you'd rather not run auth at all, Turbo's free tier is worth a look.
Marketing team with scheduling, approvals and several locales. None of the three, really. Lucky Media is clear that Keystatic isn't the right tool for editorial independence, and the same logic applies to the others at scale. That's headless CMS territory.
If you'd rather have someone wire the editor, auth and Astro collections together for a client site, that's the sort of work covered under headless CMS development.
FAQ
Can I switch editors later without migrating content?
Mostly, yes. The posts are Markdown or MDX files in your repo, so the content stays put. The work is rewriting the editor's schema so its fields match your frontmatter exactly, then re-doing auth. Keep the Astro collection schema as the source of truth and the switch is an afternoon of config, not a data migration.
Does Decap CMS work with Astro?
Yes, because it's framework-agnostic. It reads and writes the same files Astro's collections read. The trade-off, per Kahuk, is that there's little Astro-specific tooling, so you maintain the YAML config yourself.
Is Keystatic safe to use at 0.x?
It's pre-1.0, with core 0.6.9 reported as the latest release in August 2026 by GitCMS, a competitor. Pin the version, read changelogs before upgrading, and keep the content in plain files so you aren't locked in.
Which editor has scheduling and approvals?
Not Keystatic, according to Lucky Media. Tina's Editorial Workflow is described as a Team Plus feature at $49 per project per month, with scheduling and approval chains still limited. Decap documents an editorial workflow built on pull requests, though Lucky Media considers its editing UI dated.
All three keep content as files in Git, so the real switching cost is auth and the editor's schema, not your posts. The sharpest caveat is Tina's Astro visual editing, where the docs and a 2026 review disagree, so read the current Astro guide before promising it to a client.