Single binary · Windows & Linux, macOS coming soon · No Docker, guided setup
Elicana reads your documents, compiles each one into a structured knowledge note, and lets you query across everything in plain English, with local-first handling for sensitive archives.
Cancel anytime · honest beta with a direct fix line · see exactly what leaves your machine
$ elicana sources add local ~/Documents/client-archive
✓ Added · rescans every 30s, compiles in the background
$ elicana ingest https://example.com/local-first-ai
✓ Queued: "The case for local-first AI tools"
$ elicana query "Phoenix market, 2023?"
→ 12 compiled notes (2023 Q2–Q4):
· Demand strongest in multi-site health clinics
· CAC risk elevated after Q3 saturation
· Expansion via partner-led GTM onlyMost people try this first. It still leaves them with raw fragments, fragile context, and rising query cost.
Whether you upload files to Claude's project, point an AI at your folder, or connect it via MCP, the AI retrieves your raw source material: first drafts, crossed-out ideas, half-formed thoughts. It reads the pile, not your settled understanding.
No memory across conversations. No structure that persists. No connections between documents that carry forward. You re-establish context from scratch every time you open a new chat, even if you asked the same question last week.
Context windows hit their limits fast on real archives. Chunking across multiple calls compounds cost quickly. At common API pricing, querying a full archive can climb into multi-dollar-per-question territory.
Once the archive gets real, brute-force context loading stops being practical.
Point Elicana at your folder: an Obsidian vault, Google Drive sync, PDF archive, or email export. Or send a single web URL and any article you want to keep becomes part of your knowledge base instantly (coming soon).
Elicana reads each document, detects its type, and produces a structured knowledge note. Stored in Elicana's local database on your machine.
Ask in plain English. Answers are grounded in your compiled knowledge: concise, cited, and fast.
A compiled knowledge base understands conclusions, timelines, and conflicts. A retrieval tool only understands similarity.
Ask a retrieval-based tool what was concluded and it returns every document where the topic came up: every draft, every debate, every interim position. Ask the same question ten different ways and you'll get the same fragments in a different order, never the answer. Elicana compiles each document into a structured summary that preserves conclusions and key outcomes, so when you ask, the answer is already there, not scattered across a pile of raw fragments.
Retrieval tools have no concept of time or narrative arc. They surface chunks ranked by similarity with no awareness of sequence. Elicana compiles documents with date tracking, so you can ask how a position changed from January to March and get a coherent timeline, not a pile of fragments circling the same content endlessly.
When two documents assert different conclusions about the same topic, a retrieval tool surfaces both with equal confidence and no signal that they conflict. Elicana can scan your compiled notes for likely contradictions and flag them, so an answer drawn from conflicting sources tells you so (coming soon).
If you live in Obsidian, Elicana speaks your language. Point it at your vault alongside your other sources: PDFs, email, research papers, client docs, web articles. Elicana compiles what is in your vault with the material that never belonged in Obsidian in the first place. Your original vault is never touched.
Type: Deliverable · Compiled 2026-05-20 ## Summary Multi-site health clinics show strongest demand... ## Connections → [[Phoenix GTM Strategy]] → [[CAC Analysis Q3]] · [[Client: MedCore]]
Elicana reads your source files and understands your tags and wikilinks, but never writes to them. Remove Elicana any time. Your original vault is exactly as it was.
With local models, your source documents remain on-device throughout ingest, compilation, and query. See how Elicana compares to the alternatives.
Today, Elicana's compiled knowledge base is queryable through its own research interface and REST API, with per-source retrieval scores and provenance-filtered synthesis. A Model Context Protocol (MCP) server, so any MCP-compatible agent runtime can query it directly, is planned, not yet shipped.
Ask questions grounded in your personal knowledge base, not training data, not the web, but your compiled notes, through Elicana's research interface or its REST API today.
Plugging Elicana into Claude Desktop, OpenClaw, or custom agent runtimes as a native MCP tool is on the roadmap. Until then, agent workflows can call the REST API directly.
Elicana runs as a background service and resolves queries against the latest compiled state. No manual context loading, no stale retrieval.
Provenance-filtered synthesis and URL ingestion ship in the current beta, reachable via the REST API. The MCP server is planned, not yet available.
I spent two decades building products across voice, networking, and cybersecurity, starting as a software engineer before moving into product management. That leaves you with one strong instinct: be careful where sensitive data goes.
When I went independent as a consultant, the problem showed up fast. Within a few engagements, I had hundreds of files: strategy decks, research notes, deliverables, and months of meeting transcripts. I knew answers were in there. I just could not retrieve them reliably.
I tried the usual path: search, cloud AI tools, and a few self-hosted stacks. Search returned fragments. Cloud tools raised privacy concerns. Self-hosted setups asked for too much infrastructure just to work with my own notes. So I built a rough internal tool that watched a folder, compiled documents into structured notes, and let me query the result.
It changed how I prepared for calls and reused prior thinking. Then I started seeing the same frustration everywhere else. Elicana is the product version of that tool: a simpler, local-first way to turn a document archive into something you can actually use.
Ameet Kulkarni
One paid plan during the open beta, direction (not dates) for what comes next, and a separate path if you need this for a team.
The individual plan is scoped to one person. If you're a 2–25 person firm evaluating this for the team, that's a different conversation and a different price.
Contact us for pricing →Get Elicana today if you have a serious document archive and need answers you can trust. $20/month, early-bird pricing, cancel anytime.
Solo-built. No investors. Local-first by design. Elicana is built for people who need useful answers from sensitive document archives.
Not ready to buy yet? Get occasional product updates instead.