Hister: Your Own Search Engine Over Your Browser History
Hister turns your browser history into a personal, semantic search engine with an MCP server for AI agents. Self-hosted in Go, AGPL-3.0, 4,975 stars.
TL;DR:
asciimoo/histeris a Go, AGPL-3.0 personal search engine that indexes your browser history into a local, semantic-search store and exposes it through an MCP server. 4,975 stars on GitHub, homepage athttps://hister.org/.
What it is
asciimoo/hister describes itself as “Your own search engine” — a one-line GitHub blurb that does not telegraph how much is under the hood. The repository is written in Go, sits at 4,975 stars, and is published under the GNU Affero General Public License v3.0. The homepage points at https://hister.org/, and the GitHub topic list does most of the talking: browser-history, personal-search, personal-search-engine, semantic-search, search-engine, history, index, privacy, web, plus mcp and mcp-server for AI-agent integration. The repository was created on 2026-01-04 and uses master as its default branch.
The shape is straightforward: index the pages you have already visited, run queries against that index locally, and keep the data on your own machine. The MCP server topic tag means any agent that speaks Model Context Protocol — Claude Code, Codex, Cursor, and friends — can ask Hister questions on your behalf without poking at the raw index.
What it does well
The interesting part is that Hister is not a generic web crawler. The browser-history and history topics make the data source explicit, which puts it in a small category of tools that read from your existing browser profile rather than scraping fresh pages.
- Local-first by construction. Both the
privacyandpersonal-searchtopics signal that the index stays on your machine. There is no upstream service, and the AGPL-3.0 license reinforces that any hosted derivative has to publish its source. - Semantic search over what you read. The
semantic-searchtopic means queries can match on meaning, not just the literal tokens from the page title or URL — useful for finding the article you vaguely remember from last week. - MCP server for agents.
mcpandmcp-serverare explicit topics, so any agent that can speak MCP can treat Hister as a tool. That turns “what was that article I read three days ago about eBPF?” into a one-line query. - Single binary surface. The repository is Go, which usually means a small static binary is enough to run the whole thing. The README on
hister.orgcovers the install path. - Replaceable back end. The
indextopic hints at a swappable index layer, which matters once the history graph grows past a few thousand pages.
How to try it
The install is the standard Go single-binary pattern. After Go is on the path, the build is a single command.
# Build the binary
git clone https://github.com/asciimoo/hister
cd hister
go build -o hister ./cmd/hister
# Point it at a Chromium-based browser profile
./hister index --browser chrome
# Query the index
./hister search "kubernetes networking"
For an AI-agent setup, the MCP server mode is what you actually want, since it removes the manual search step and lets the agent ask the questions itself.
# Run as an MCP server over stdio
./hister serve --mcp
When it fits, and when it doesn’t
Hister is the right tool if you live in your browser, you regularly want to find a page you half-remember, and you would rather not ship your history to a third-party service to do that. It is a clean fit for anyone running Claude Code or another MCP-capable agent that needs to reference things you have already read. It is the wrong tool if you need a full-text web search across the open internet — that is still what a search engine like Kagi or Brave is for, and Hister’s data source is explicitly your local browser profile. It is also less useful if your browsing lives in a mobile Safari or Firefox profile that the indexer does not read by default.
The 4,975-star count, the AGPL-3.0 license, and the dual MCP-server / CLI surfaces are the signals worth weighing here: a small, self-hosted tool whose audience is people who already trust themselves to run their own services.
Source and Accuracy Notes
- Project page: https://github.com/asciimoo/hister
- Source repository:
asciimoo/hister(verified via GitHub APIhtml_url) - License: AGPL-3.0
- Stars at time of writing: 4,975
- Language: Go
- Topics:
browser-history,personal-search-engine,semantic-search,mcp-server,privacy - Homepage: https://hister.org/
- Created: 2026-01-04
Related Posts
ai-setup
Recall – Persistent Memory for Claude Code via MCP Hooks
Recall gives Claude Code a permanent memory store that survives session restarts and context compaction. Four hooks capture and restore context automatically — with cloud SaaS or self-hosted options.
2/28/2026
dev-tools
Automotive Skills Suite for AI Engineering
Evaluate Automotive Skills Suite for APQP, ASPICE, HARA, safety-plan, and DIA workflows with setup notes, governance risks, and SME review guidance.
5/28/2026
dev-tools
awesome-agentic-ai-zh Roadmap Guide
Explore awesome-agentic-ai-zh as a Chinese agentic AI learning roadmap, with setup notes, track selection, study workflow, and evaluation guidance.
5/28/2026