The knowledge base for the AI era

Where knowledge goes to live.

A markdown wiki your team actually enjoys — wired so your AI agents can read, search, and reason over everything you know. Same source of truth. Two kinds of readers.

Explore the live wiki → Learn more

Be honest: nobody's opened your wiki since 2019.

It's stale, the search is a slot machine, and the knowledge that matters lives in someone's DMs. Now your shiny AI assistant can't use any of it either. Two problems, same root cause: your wiki was built for nobody in particular.

Why Wikantik

Four things your current wiki doesn't do

Markdown in, freedom out

Plain markdown, git-backed, self-hostable. Your knowledge is yours — export it, diff it, walk away with it. No lock-in, no hostage situation.

Search that actually finds it

Hybrid retrieval — keyword + meaning + a knowledge-graph rerank. Ask in plain language, get the right page. Not keyword roulette.

Your AI's favorite coworker

Native MCP servers mean your agents query, retrieve, and cite your wiki directly. The knowledge graph hands them entities and relationships, not a wall of HTML.

Built for two readers

Every page serves humans (clean, fast reading) and machines (token-budgeted projections, runbooks, verification metadata). Write once; both get the good version.

Platform

An engine your AI agents can actually use

Native MCP servers, hybrid retrieval, and a knowledge graph of your own content — the machinery that makes Wikantik answer-engine ready.

MCP for AI agents

Your agents query, retrieve, and cite your wiki over the Model Context Protocol.

Hybrid retrieval

BM25 + dense embeddings + a knowledge-graph rerank. Ask in plain language.

Knowledge graph

Entities and relationships extracted from your pages, not a wall of HTML.

Ontology & SPARQL

Your whole wiki as queryable linked data — SPARQL, dereferenceable IRIs, a SHACL-governed model.

Explore the platform →

Enterprise

Ready for the security review

SSO, SCIM provisioning, a tamper-evident audit log, role-based access control, and self-hosting with your own backups. Own your data; pass the questionnaire.

SSO (SAML / OIDC)

SCIM provisioning

Automated onboarding and offboarding driven by your identity provider.

Tamper-evident audit

A verifiable, exportable trail of every privileged action.

See enterprise features →

Ask your wiki. Out loud. Like a person.

Your AI assistant talks to Wikantik over MCP — and answers from your knowledge, with citations.

→ "How do we roll back a bad prod deploy?"

✓ Found runbook: Rollback Procedure (verified 3d ago)

1. bin/remote.sh rollback — re-promotes the :rollback image

2. DB + pages persist across the swap …

cited 2 pages · 1 knowledge-graph entity

How we're different

(We'll keep it sporting. Mostly.)

Capability	Wikantik	Confluence / Notion	MediaWiki / Obsidian
Plain markdown, you own your data	✅	⚠️	✅
Hybrid AI search built in	✅	⚠️	❌
Agents read it natively (MCP)	✅	❌	❌
Knowledge graph of your content	✅	❌	❌

Hosting

Run it your way. Or don't run it at all.

Wikantik is yours either way. Host it yourself, or let us keep it alive while you get on with your life.

Self-managed

Docker up, point it at Postgres, done. Runs on your infra, your rules. Open and yours forever — no asterisks, no seat counter.

Free. It's just software.

Easiest

Fully managed

We run it, patch it, back it up, and keep it fast. You write pages and point your agents at it.

Priced on compute, not per seat — lightly marked up. Add your whole team for free.

Learn more →

Curious? Let's talk.

Tell us a bit about your team and we'll show you how to get Wikantik running — your way or ours.