Your Browser's Memory Is Broken

Modern browsers offer a broken promise of memory. Your history records only two things: a page's title and its URL. This fundamental flaw means the actual content you read, the substance you remember, vanishes the moment you close the tab, leaving it nearly impossible to retrieve unless you recall the exact headline. Firefox's default history expiry of 90 days only compounds the problem.

This limitation highlights a crucial distinction in information retrieval. discovery search — finding new information — is Google’s undisputed domain. But recall search — locating something you’ve already encountered — remains a largely unsolved problem, forcing us to awkwardly "Google our way back" to pages we know we've seen.

Enter Hister, an open-source, self-hosted solution that finally fixes this. It quietly indexes the full text of every page you visit, operating like a private search engine for your personal web history. A browser extension for Chrome and Firefox automatically captures rendered page content, making your browsing history genuinely useful for recall, not just a list of forgotten links.

Setting up your personal search engine takes minutes. Download the appropriate Hister binary for your OS (e.g., Darwin ARM64 for Mac M4), make it executable with chmod +x hister, and run hister listen. This command starts a local server, typically on port 4433, ready to receive data. The accompanying lightweight browser extension for Chrome or Firefox acts as your silent data collector, forwarding content to your local Hister server as you browse.

Indexing happens entirely in the background, without any manual intervention. Unlike read-it-later services requiring conscious "save" actions, Hister's extension automatically captures the full text of every page you visit. It builds your comprehensive index as you browse normally, eliminating the friction of manual curation. This passive collection ensures nothing important slips through the cracks.

Witness the power of full-text recall with a simple search. Imagine trying to find an article you read weeks ago, remembering only a random sentence from its body, not the title or URL. Standard browser history utterly fails this. Hister, however, instantly retrieves the exact page, highlighting the matching text. This capability extends to internal dashboards or private repos—content Google can never index—turning your browsing history into a powerful, searchable corpus.

Beyond simple keywords, Hister offers a sophisticated query language, enabling surgical precision in your personal recall. Filter results with commands like domain:github.com to target specific sources, or text:memory leak to pinpoint exact phrases within page content. This powerful syntax transforms vague memories into actionable searches, letting you zero in on information you've read.

Hister builds its comprehensive knowledge base from multiple sources, far beyond just new browser visits. It can import your entire existing browser history, pulling in past URLs for re-indexing. Furthermore, it watches local file directories, integrating personal documents and code into your searchable index. The system even crawls entire websites, transforming disparate data into a unified, private search engine for recall search.

Most forward-looking is Hister's MCP endpoint. This unique feature allows AI coding agents to connect directly to your private index. Agents ground their responses in documents you have read and authenticated, not just generic models or disparate external sites. This provides contextually relevant and personalized assistance, leveraging your curated knowledge. Explore its capabilities further on asciimoo/hister: Your own search engine - GitHub.

Hister, while powerful, arrives with caveats. The index, containing the full text of everything you’ve read, remains unencrypted by default; full disk encryption is essential. Its default Docker Compose file also binds to 0.0.0.0 with no authentication, presenting a significant security risk that demands immediate configuration changes.

Safari users are out of luck; despite a proof-of-concept pull request, support remains absent. Mac users downloading the binary will also encounter Gatekeeper warnings as it’s ad-hoc signed, not notarized. These are sharp edges for a tool designed for deep personal data.

Project currently sits pre-1.0, with an active but often breaking development cycle. Expect changes: 18 minor releases in seven months included documented breaking changes. While 47 contributors are listed, roughly 78% of contributions come from the primary maintainer, highlighting a focused, but centralized, effort.

So, who is Hister for now? Developers, researchers, and seasoned self-hosters on Chrome or Firefox. It’s a compelling tool if you navigate complex documentation, GitHub issues, and value a private corpus for your coding agents. If you require polished stability or lack comfort with technical configuration, consider waiting until it matures past its pre-1.0 status.

What is Hister?

Hister is an open-source, self-hosted tool that creates a private, full-text search engine out of your browsing history. It indexes the content of every page you visit, allowing you to find information based on what you remember reading, not just the page title or URL.

How does Hister protect my privacy?

Hister runs entirely on your local machine. The browser extension sends data only to a local server you control. All your data, including the full text of pages you've visited, is stored on your own disk, ensuring it never leaves your system.

Does Hister work on all browsers?

Hister has official extensions for Chrome and Firefox. It does not currently work with Safari, as a proof-of-concept pull request has been inactive for months.

Can Hister search my old browser history?

Yes, Hister can import your existing history from Chrome or Firefox. However, since your browser didn't save the original page content, Hister must revisit those URLs to index whatever content is currently live on those pages.

What is the MCP endpoint in Hister?

The MCP (Multi-Client Protocol) endpoint allows AI agents or copilots to connect to your Hister index. This enables an AI to answer questions using the private corpus of everything you've personally read, rather than its general training data.