Slovak legislation, searchable

Search Slovak law in plain language — from your browser, from any HTTP client, or from an AI assistant over MCP. Free, and no account needed.

Try it Connect it Source on GitLab soon

The code is private while we finish a security review — we're working hard on making it public. Tell us if you want a heads-up.

documents parsed · searchable so far ·

About

Law lookup,
not a legal opinion.

parses the full text of Slovak legislation into a searchable tree of sections, and finds the ones relevant to a question asked in plain language. Every section has a stable link, so an answer can cite the law it came from.

Searching is free and needs no account. We embed your query on our own hardware, which is modest — so at busy times the free tier is best effort. If you would rather not wait, you can embed the query yourself and send us the vector; that path skips our queue entirely, because it costs us nothing but the lookup.

No account needed Search anonymously over HTTP or MCP. Nothing to sign up for.
Bring your own embeddings Send a query vector instead of text and skip the queue. Optional.
Citable sections Every paragraph has a permanent link, for humans and assistants alike.

Indexing is still in progress. All documents are parsed and browsable, but only are semantically searchable so far. The rest are being embedded. The source lives on GitLab and is private for now — we're working hard on making it public, once the indexing and a security review are done.

Setup

Use it however
suits you.

No client, no account, no key. Post a question, get the matching sections:

curl -X POST  \
  -H 'Content-Type: application/json' \
  -d '{"query": "Aké sú podmienky výpovede?", "limit": 5}'

Each result carries a node_id and a permanent link at /doc/{document_id}?node={node_id}.

Add the following to your Claude Desktop config (claude_desktop_config.json):

{
  "mcpServers": {
    "lex-index": {
      "url": ""
    }
  }
}

Restart Claude Desktop and you're ready. No credentials required.

Any client that speaks remote MCP over HTTP can connect to:

Support for remote MCP servers is still uneven across assistants, and setup differs between them. If yours cannot connect, the HTTP API above works everywhere. We are building a small client of our own so that connecting does not depend on your assistant catching up.

Embed the query yourself and send the vector instead of the text. This skips our embedding queue, because the only work left for us is the lookup: 

curl -X POST  \
  -H 'Content-Type: application/json' \
  -d '{"query": "…", "embedding": [0.01, …], "embedding_type": "nomic"}'

The vector must come from the same model as the index you search — currently nomic-embed-text (768 dimensions). A vector from a different model will return confident nonsense rather than an error.

Try it

See what it finds.

Searching

This is a preview of what your AI assistant receives as context.

Plans

Free, and
meant to stay that way.

Searching Slovak law should not cost anything. What costs us money is turning your question into a vector — so that is the only thing anyone would ever pay for, and only if they want to skip the line.

Available now

Free

€0no account

  • Full corpus search, over HTTP or MCP
  • We embed your query for you
  • Best effort — shared
  • Permanent links to every section

Available now

Free, your embeddings

€0no account

  • Everything in Free
  • Send a query vector instead of text
  • Skips the embedding queue — it costs us only the lookup
  • You choose the model, and run it

In development

Priority

—pricing not set

  • Everything in Free
  • Skip the queue without running a model
  • Higher-quality embeddings for better recall
  • Pay for what you use, top up when it runs out
Tell us you want this →

Available now

Dedicated

Hardwareat cost

  • Your own instance and database
  • Guaranteed capacity, no shared queue
  • Run whatever you like against it
  • For teams that outgrow the free tier
Get in touch →

There is no signup and no card on file, because there is nothing to buy yet. Priority needs accounts, metering and billing before it is real — we would rather ship it late than charge for a queue that does not exist.

Support

Free and community-funded.

is free to use. It runs on a single Hetzner instance that we pay for out of pocket, and that box is the whole capacity.

Instance

Monthly cost

Embedding throughput

That last number is the honest limit: queries we embed for you are capped by this hardware. Embedding your own query removes you from that bottleneck.

If you find useful and want to make it faster for everyone, you can contribute to the hardware budget pool.

If you're an individual or company that needs a dedicated instance with guaranteed capacity, get in touch.

Contribute to HW budget Request dedicated instance