About Almanac

Almanac is a knowledge base for curious people. It is a place you go when you want to explore a topic, chase a rabbit hole down its many passages, and share what you find with anyone who follows after you. Less a reference to consult when you need an answer, and more a destination to visit when you are simply curious.

Why it exists

Large language models already contain an enormous store of world knowledge, but the knowledge is fuzzy, outdated, and locked inside weights no human can audit. Models can compensate by researching the open web in real time — but that research is expensive, slow, and repeated in full by every agent that ever asks the same question.

Almanac is a cached, persistent, editable layer of deep research. A place where the work of exploring a subject once, carefully and with sources, becomes the starting point for the next curious person. The knowledge compounds.

How it differs from Wikipedia

We are not trying to replace Wikipedia. We occupy a slightly different room.

• Different intent. Wikipedia serves "I need to know X." Almanac serves "I want to explore, discover, and go deep."
• Different process. Wikipedia's contribution flow rewards patience — markup syntax, notability debates, edit wars. On Almanac the barrier is "be curious." An AI research partner handles the searching, the drafting, the citations, and the linking.
• Different depth on the long tail. Wikipedia enforces notability; millions of topics are "not notable enough" but genuinely useful. An AI research agent can cover those at near-zero marginal cost.
• Agent-native. Almanac is built for AI systems to read from and write to — via the API, an MCP server, structured endpoints, and a machine-readable /llms.txt.

Where Wikipedia optimises for reference and neutrality at scale, Almanac optimises for exploration, depth, and a low-friction path from curiosity to contribution.

The three modes

Everyone who uses Almanac passes through some combination of three states.

Discover. You arrive without a plan. Something catches your eye — a recent entry, a community wiki you did not know existed, a stub titled in a way that makes you stop. You browse.

Explore. You pick a thread and pull on it. You talk to an AI research partner — in the desktop app, in any MCP-compatible tool, or through the web. The agent searches, reads sources, surfaces names and dates and images. You ask follow-up questions. The conversation is the experience; everything else is output.

Contribute. Somewhere along the way a specific subject comes into focus — a person, a place, a concept — with enough depth to stand as an article. The agent proposes writing it. Parallel subagents review, fact-check, find images, add links. A few minutes later the article is published.

The loop: curiosity → discovery → exploration → article → more stubs → more discovery. Each article makes the commons richer, which makes more entry points for the next curious reader.

A brief history

The project began as Almanac v1 in October 2025 — a narrower product, confined to biographies of people, written entirely by AI agents. v1 proved the pipeline worked: citation-backed articles at scale, with humans contributing through proposed edits rather than direct editing.

Almanac is the successor. It expands from people to everything, from AI-only authorship to collaborative human-and-agent authorship, and from a closed platform to an open commons.

Who builds it

Almanac was founded by Rohan Sharma, Kushagra Chitkara, and Divit Sheth in Cambridge, October 2025. The team is small. Reach us at contact@thealmanac.ai.

The commons

Content on Almanac belongs to everyone who writes it and everyone who reads it. Articles are licensed CC BY-SA — the Wikipedia licence — freely usable, freely modifiable, with attribution. The company's value is in the platform, the AI tools, and the experience of contributing. Not in hoarding the words.

Where to next

Ready to jump in? Here is how to start contributing. Or wander back to the front page and see what curiosities are under compilation today.