{"slug": "hister-self-hosted-search-engine-for-your-browser-history", "title": "Hister: Self-Hosted Search Engine for Your Browser History", "summary": "Hister, an open-source self-hosted search engine that indexes the full text of every page a user visits and every local file they keep, hit number one on Hacker News this week with over 650 upvotes. Hister installs via Homebrew with `brew install hister` and `hister listen`, runs at http://127.0.0.1:4433, and exposes a Model Context Protocol server at /api/mcp so AI coding agents such as Claude Code can query the user's browsing history. The tool backfills existing history with `hister import browser` from Chrome, Firefox, or Safari on macOS, and its query syntax supports filters including title:, domain:, visits:5.., added:>30d, and type:file.", "body_md": "You read an article three weeks ago about Kubernetes admission webhooks. You can’t remember the URL. Your browser history shows a title — if you scroll to the right day. You search Google, find the topic, but not *that* explanation. Twenty minutes gone. [Hister](https://hister.org/) exists to end this. It’s an open-source, self-hosted search engine that indexes the full text of every page you visit and every file you keep — and it ships an MCP server so your AI coding agent can query all of it. It hit number one on Hacker News this week with over 650 upvotes and deserves the attention.\n\n## What Browser History Gets Wrong\n\nChrome and Firefox store a URL and a page title. That’s it. You can’t search by what a page *said*. You can’t filter by domain, visit frequency, or date range with any precision. If you read five tabs about Rust async runtimes and closed them, they’re essentially gone — findable only if you remember the exact site.\n\nHister indexes the actual content. Install the browser extension for Chrome or Firefox, run the server locally, and every page you visit gets its full text captured and stored in a searchable index on your own machine. No cloud. No subscription. No one reading your browsing data.\n\n## Setup in Two Steps\n\nThere is no complex configuration required to start. Install via Homebrew and launch the server:\n\n```\nbrew install hister\nhister listen\n```\n\nOr download a prebuilt binary from the [open-source GitHub repository](https://github.com/asciimoo/hister). The server starts at `http://127.0.0.1:4433`. Install the browser extension. Browse. Search. That’s the entire onboarding path.\n\nFor existing history, `hister import browser` backfills from Chrome, Firefox, or Safari on macOS. You can also point Hister at a local folder to index files, or run a crawler against a documentation site you reference often.\n\n## The MCP Endpoint Changes the Value Proposition\n\nHere is where Hister goes from “useful browser tool” to meaningful infrastructure. Hister exposes a [Model Context Protocol](https://modelcontextprotocol.io/) server at `/api/mcp`. Wire Claude Code to it with four lines of config:\n\n```\n{\n  \"mcpServers\": {\n    \"hister\": {\n      \"url\": \"http://localhost:4433/api/mcp\"\n    }\n  }\n}\n```\n\nYour agent can now search everything you’ve read. Ask Claude Code about a caching pattern and it can pull the articles you’ve bookmarked on Redis TTL strategies without you pasting URLs. This is RAG over your personal knowledge base — no vector database, no API bill, no cloud pipeline. The official [Hister MCP guide](https://hister.org/posts/give-your-ai-assistant-a-private-memory) walks through the integration, and the [MCP documentation](https://hister.org/docs/mcp) covers token configuration for server deployments.\n\n## Query Syntax Worth Knowing\n\nSimple keyword search works as you’d expect. But the query language has real depth:\n\n- `title:kubernetes admission` — search page titles only\n- `domain:docs.rust-lang.org` — restrict results to a domain\n- `visits:5..` — pages visited five or more times\n- `added:>30d` — indexed in the last 30 days\n- `type:file` — local files only\n\nCombine them. `title:postgres domain:github.com visits:3..` surfaces GitHub pages about Postgres you’ve returned to repeatedly — the issues and PRs that actually mattered, not every result you ever clicked.\n\n## How It Compares to Rewind and Obsidian\n\nThe obvious comparison is [Rewind AI](https://www.rewind.ai/). Rewind captures your screen continuously, processes it in the cloud, costs $30 a month, and runs only on macOS. Hister captures page text, processes it locally, costs nothing, and runs on anything. The tradeoff: Rewind records everything including desktop apps. Hister covers what you browse and what you explicitly add. For developers who spend most of their day in a browser and a terminal, Hister’s coverage is sufficient.\n\nObsidian with a web clipper is another common answer to this problem. The difference is friction. A web clipper requires you to remember to clip. Hister requires nothing after the initial setup. Three months of passive accumulation is what makes it valuable.\n\n## The Honest Limitations\n\nHister is at version 0.19.0. Some rough edges worth knowing before you deploy:\n\n- Single-page apps that lazy-render with React or Next.js may not capture cleanly if the extension fires before content renders\n- Login-walled pages only get indexed when you’re authenticated during the visit\n- Semantic search requires a separately configured embeddings endpoint — Ollama works, but it is not bundled\n\nNone of these are dealbreakers for the core use case. The SPA rendering issue is on the roadmap, and the [Hacker News thread](https://news.ycombinator.com/item?id=49351802) has practical workarounds for several edge cases.\n\n## Worth Deploying This Weekend\n\nThe passive browsing history angle is useful enough on its own. The MCP integration is what makes Hister worth deploying now rather than waiting for a more polished release. If you run Claude Code or Cursor, giving your agent access to years of your reading history is a concrete capability upgrade. Setup takes ten minutes. The payoff compounds.", "url": "https://wpnews.pro/news/hister-self-hosted-search-engine-for-your-browser-history", "canonical_source": "https://byteiota.com/hister-self-hosted-search-engine-browser-history/", "published_at": "2026-09-19 10:13:10+00:00", "updated_at": "2026-09-19 10:24:11.092516+00:00", "lang": "en", "topics": ["ai-agents", "agent-protocols", "ai-tools", "ai-search", "artificial-intelligence"], "entities": ["Hister", "Hacker News", "Model Context Protocol", "Claude Code", "Chrome", "Firefox", "Safari", "Rewind AI"], "alternates": {"html": "https://wpnews.pro/news/hister-self-hosted-search-engine-for-your-browser-history", "markdown": "https://wpnews.pro/news/hister-self-hosted-search-engine-for-your-browser-history.md", "text": "https://wpnews.pro/news/hister-self-hosted-search-engine-for-your-browser-history.txt", "jsonld": "https://wpnews.pro/news/hister-self-hosted-search-engine-for-your-browser-history.jsonld"}}