Vertex vs Karakeep 2026: New Wave of Personal Archivers

On August 27, 2026, a Show HN post titled "Vertex, your personal archiver and annotator" appeared on Hacker News with a description that reads almost like a reply to the Internet Archive: web pages change, link-rot creeps in, and a personal vault of annotated pages plus immutable snapshots is the answer. Four days earlier, Karakeep (formerly Hoarder) crossed 28,684 GitHub stars with the same proposition from the opposite direction: self-hosted, AGPL-3.0, with AI tagging baked in. Together, the two launches bookend a 2026 wave of personal archivers that has been quietly reshaping what "archive" means for individuals. Public social archives like ThreadGrab sit on the other side of the same trend.

What Vertex actually is

Vertex is a Chrome browser extension plus a web app, currently Chrome-only with Firefox on the roadmap. The product is built in Haskell on the backend and TypeScript on the frontend. The author described the motivation in the HN post in the same terms every archiving project uses: "links I cared about silently changed or disappeared, and once my collection got large enough, I lost track of which identical-looking URL I wanted."

The model has three primitives:

• Vertex — a saved bookmark plus optional notes, set membership, and graph edges.
• Snapshot — a self-contained MHTML document stored immutably, independent of the source page.
• Edge — a typed relationship between two vertices that records how the user perceives them to be related.

The Chrome extension surfaces saved links with their notes inline as the user browses, so the "did I already save this?" question becomes visible at the moment it matters. Pricing splits into three tiers:

The pricing choice is itself a position. Vertex is selling snapshots as a paid utility while keeping the annotation graph free. The implied audience is the working researcher or knowledge worker who is willing to pay for guaranteed retrieval but does not want their bookmarks locked behind a paywall.

What Karakeep actually is

Karakeep is the open-source incumbent in the same niche. The repository, github.com/karakeep-app/karakeep, is licensed AGPL-3.0, last updated August 31, 2026, and was created on February 6, 2024. The headline feature list is bookmarking, notes, images, AI-based automatic tagging, and full-text search.

The capture side leans on SingleFile (a popular browser extension that saves complete pages as a single HTML file) plus native browser APIs. Self-hosters run it via Docker; non-hosters pay about $4 per month for the hosted version per a comment on the Vertex HN post by user toomuchtodo:

Check out github.com/karakeep-app/karakeep for inspiration. Its extension has SingleFile built in for browser capture and costs $4/month if you don't want to host it yourself. Great work!

Karakeep and Vertex disagree on three points and agree on the rest. The disagreements matter when you choose between them:

The pattern: both projects believe link-rot is the central problem; both bundle the captured page with the bookmark; both expose the result as a searchable collection. The difference is whether the user owns the storage stack (Karakeep) or pays the project to own it (Vertex).

Why the new wave is happening now

Three forces converged to make August 2026 a personal-archiving inflection point.

1. Link rot has caught up with how people actually work. Vertex's landing page states the canonical claim: "Link-rot affects over half of all URLs within a decade as pages get updated, paywalled, or taken down without notice." This is the same number every archiving project has cited since the Internet Archive's own studies in 2014. What changed is that knowledge workers now do a meaningful share of their reading inside ephemeral surfaces (X threads, Substack posts, Bluesky long-form, LinkedIn newsletters) where the source can vanish within weeks, not years.

2. The capture primitives have become cheap. Browser extensions like SingleFile, Chrome's own Page Capture API, and CF Browser Run (the underlying engine behind ThreadGrab's headless capture) make it trivial to render and snapshot a page on demand. The remaining question is storage and retrieval, which is exactly where Vertex and Karakeep compete.

3. Self-hosting is back in fashion. Karakeep's 28,684-star trajectory and the AGPL-3.0 license reflect a community that is comfortable running Docker on a $5 VPS or a home server. Vertex's cloud-first model bets the other half of the market wants to pay a SaaS bill rather than operate infrastructure. Both audiences are large enough to sustain their respective products.

The four quadrants of the 2026 archive market

Once you place Vertex and Karakeep on a chart, the broader market becomes clear. Two axes matter: who owns the storage (you vs the service) and whether the archive is public or private.

Vertex and Karakeep both occupy the private-vault row. ThreadGrab occupies the public-archive row on the cloud-hosted side. The reason this quadrant map matters: a creator's full archive stack usually combines one vault tool with one public tool, not two competitors from the same row.

Where ThreadGrab fits

ThreadGrab turns X threads, Bluesky long-form posts, and LinkedIn newsletters into structured Markdown that is readable, citable, and SEO-friendly. The positioning is the mirror image of Vertex and Karakeep:

• The unit is the post, not the page. ThreadGrab extracts structured fields (author, date, thread id, media URLs, quoted posts) rather than archiving the rendered page. Vertex and Karakeep treat the rendered page as the unit.
• The reader is the public, not the saver. Anyone with a ThreadGrab URL can read the thread. Vertex and Karakeep keep the saved page inside the user's account.
• The output is Markdown, not HTML or MHTML. ThreadGrab loses layout on purpose so the content can be re-rendered in any reader, in any front-end, in any language.
• Retention is public and durable, not user-bound. A ThreadGrab post stays accessible because it lives on a public CDN, not because the original saver paid their subscription this month.

The complementary use case is concrete. A creator who writes a long Bluesky thread about, say, a programming pattern, might:

1. Save the source articles they cited in Karakeep so the research trail is private.
2. Publish the resulting thread through ThreadGrab so anyone can read, share and cite it.
3. Use Vertex to grab the public ThreadGrab URL plus any reply threads once the conversation settles.

None of those three tools overlap with the others on output or audience. The 2026 personal-archiver wave validates that the underlying demand — durable copies of mutable pages — is broad enough to support multiple specialized implementations.

A concrete code example: posting a ThreadGrab archive from a Vertex-saved source

If you maintain a public archive site alongside a personal Karakeep or Vertex vault, the workflow looks like this in practice. The example below uses the ThreadGrab API to publish a Bluesky thread that you first captured into your vault:

# 1. Save the original Bluesky thread URL into your personal vault
# (Karakeep via API; Vertex via extension right-click)
curl -X POST https://karakeep.example.com/api/v1/bookmarks \
  -H "Authorization: Bearer $KARAKEEP_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
        "type": "link",
        "url": "https://bsky.app/profile/jane.dev/post/3ltz4abc1234",
        "title": "Jane on Bluesky: a thread about TypeScript inference",
        "tags": ["typescript", "to-publish"]
      }'

Then trigger the public archive from the same URL a few hours later:

# 2. Submit the same URL to ThreadGrab for public archiving
curl -X POST https://threadgrab.com/api/v1/capture \
  -H "Authorization: Bearer $THREADGRAB_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
        "url": "https://bsky.app/profile/jane.dev/post/3ltz4abc1234",
        "lang": "en",
        "include_quotes": true,
        "include_replies": false,
        "output_format": "markdown"
      }'

# Response: {"archive_id": "tg_2026_0831_a8f2e1", "url": "https://threadgrab.com/en/blog/..."}

The two archives now live in two different quadrants: the personal one is private to you, the public one is durable for any reader. If the Bluesky post disappears, the ThreadGrab copy still serves it.

What to watch in the next 90 days

Three near-term signals will tell you whether this new wave is durable or another HN burst:

1. Vertex's Firefox extension. A second browser extension would push Vertex past the Chrome-only constraint and broaden the addressable market. The Vertex roadmap explicitly lists Firefox.
2. Karakeep's federation story. Self-hosted Karakeep instances do not share vertices today. If the project ships an ActivityPub or AT Protocol bridge, the AGPL community gets a free network effect that Vertex cannot replicate without owning the same protocol.
3. MHTML as a portability standard. Both projects rely on MHTML (Vertex) or SingleFile HTML (Karakeep). If a portability spec emerges — a shared index of "snapshots are exchangeable between Vertex and Karakeep" — the personal-vault quadrant becomes interoperable in the way email clients are.

None of those signals are guaranteed. The safer bet is that personal archivers will keep splitting along the cloud / self-host line, while public archives like ThreadGrab continue to expand the long tail of social content that is worth preserving for any reader, not just the original author.

FAQ

Last verified: August 31, 2026. Vertex pricing and Karakeep star count verified via direct vendor sites on the publish date.