Almanac

Almanac is a hosted wiki and source library that turns the files you already have into a knowledge base both people and AI agents can use. You point it at a folder of documents, and it organizes them into readable pages backed by the original sources and citations. The tagline captures the intent well, since it's a shared wiki for you and your agent. The problem it goes after is that most teams already have the knowledge they need, it just lives scattered across markdown files, spreadsheets, and PDFs, in a shape that neither a new teammate nor an agent can quickly make sense of. Almanac's job is to take that pile and give it structure, so the same material becomes something a person can read and a program can query without a lot of preparation first.

The workflow starts with a folder. You add your files, and Almanac runs a managed job that reads them and writes a source-cited wiki out of what's there. Markdown, CSVs, and PDFs all go in, and they come out organized into three connected pieces, the original sources, the pages that organize the content, and the citations that link each claim back to where it came from. That structure is what makes the wiki trustworthy, because a reader or an agent can always trace a statement to the document it rests on instead of guessing whether it's accurate. It's a different model from dumping files into a folder and hoping search finds the right one, since the pages are actually written and organized rather than just indexed, and the citations stay attached as the material grows.

Where Almanac separates itself from an ordinary wiki is that it's built for agents from the start. You can browse and edit the wiki yourself in the browser, and you can give the agents you already use access to the same workspace through a CLI, an MCP server, an SDK, or an API. That means an agent can search the wiki, read the pages, and update content, working from the same source-backed context a person would. Instead of pasting documents into a prompt and hoping for the best, the agent pulls from an organized, cited knowledge base that stays current as the underlying files change. Offering four separate paths in, from a command-line tool to a standard protocol, means most existing setups have a way to connect without rebuilding how the agent already works.

It fits teams and individuals who want their scattered documents turned into something both a person and an agent can rely on. That includes engineering and research teams building agent workflows, operators who want an internal knowledge base their tools can query, and anyone tired of an agent hallucinating because it never had the right context to begin with. Because the whole thing is hosted, there's no wiki server to run or maintain, and no retrieval pipeline to stand up yourself. You bring the files, and Almanac keeps the organized version available to whatever is reading it, whether that's a teammate in a browser tab or an agent hitting the API in the background.

Plenty of tools store documents, and plenty of tools do vector search over text. Almanac's angle is the combination of a human-readable wiki and an agent-ready interface over the same source-cited content, so the version a teammate reads and the version an agent queries are one and the same rather than two copies that drift apart. The insistence on sources and citations is the other differentiator, since it keeps the knowledge base honest and lets both people and software check the evidence behind a page rather than trusting a summary that may have wandered from the original. That shared, cited surface is the thing that's genuinely hard to assemble by gluing a wiki and a vector store together yourself.

Almanac is a hosted product, so you sign in to get started rather than installing anything, and it's clearly an early-stage tool that's still filling out its public details. Pricing isn't posted on the marketing site at the moment, and the account flow sits behind a sign-in, so the current plans are best confirmed directly with the team rather than assumed. The founder is reachable at rohan@usealmanac.com, which is a good sign for anyone who wants to ask about access, limits, or how the agent integrations would fit their setup before committing to it. For a young product, having a real person to talk to counts for something when you're deciding whether to build on it.

For anyone building with agents right now, the pitch is straightforward. Agents are only as good as the context you give them, and most of that context is trapped in files that were never organized for retrieval in the first place. Almanac tries to be the layer that fixes that, turning a messy folder into a cited wiki that a person can read and an agent can query through the same set of interfaces. If that's the gap you're hitting, it's worth a look, with the honest caveat that it's young and you'll want to talk to the team about exactly where it stands today before you lean on it heavily.