# com.valksor/livonian

> Grounds LLMs in the real Livonian dictionary: attested words, inflections, romanization.

- name: com.valksor/livonian
- version: 2.1.0
- connection class: R0 (autonomous)
- trust: 75/100 flags: no-repository
- quality: 88/100 (good)
- owner: not claimed
- descriptor: https://api.protogrid.dev/v1/servers/com.valksor%2Flivonian
- tools: https://api.protogrid.dev/v1/servers/com.valksor%2Flivonian/tools

**An agent can connect right now, no human step.** Remote endpoint, no authentication, reachable on the last probe.

## Connection (mcpServers)

```json
{
  "mcpServers": {
    "livonian": {
      "type": "http",
      "url": "https://livonian.valksor.com/mcp"
    }
  }
}
```


## Trust

- hygiene: 80
- liveness: 99
- freshness: 65
- provenance: 55
- drivers: +dns-namespace +website +reachable +uptime-97% ~updated-105d-ago -no-repository
- Derived from observable signals (official registry feed and our own credential-free probes); no code audit performed.

## Quality 88/100 (good)

### protocol: 75 (weight 25)

- warn `protocol.modern`: newest supported version is 2025-11-25; 2026-07-28 not supported, older versions are deprecated until 2027-07-28
- n/a `protocol.stateless`: only applies to 2026-07-28 servers
- pass `protocol.transport`: streamable HTTP
- n/a `protocol.list_ttl`: only applies to 2026-07-28 servers

### authorization: n/a (weight 15)

- n/a `auth.prm`, `auth.as_metadata`, `auth.cimd`: no remote uses OAuth
- n/a `auth.secret_in_url`: no templated URL

### tool hygiene: 93 (weight 30)

- pass `tools.descriptions`: every tool has a description
- warn `tools.description_length`: 1 tools have descriptions over 1,024 characters, which crowds the context
- pass `tools.schemas`: every tool has an object input schema
- pass `tools.annotations`: 4 of 4 tools declare readOnlyHint or destructiveHint
- pass `tools.directory_hints`: every tool declares readOnlyHint, destructiveHint, idempotentHint and openWorldHint
- pass `tools.token_cost`: about 2,020 tokens to load every tool
- pass `tools.api_dump`: tools are not a one-to-one API dump

### stability: 100 (weight 15)

- pass `stability.changes`: 2 tool changes in 30 days
- pass `stability.rug_pull`: no tool changed its meaning under the same name

### dependencies: n/a (weight 15)

- n/a `deps.known_vulns`, `deps.mcp_sdk_version`, `deps.resolvable`: no npm or PyPI package

- tools: 4, about 2,020 tokens to load them all
- tool set hash: 19d193468eebdce1ea4d40cde28d36a6
- tools last changed: 2026-10-13T12:10:41.582Z
- badge: [![protogrid quality](https://protogrid.dev/badge/com.valksor/livonian.svg)](https://protogrid.dev/servers/com.valksor/livonian)
- Checks run on what our credential-free, read-only probes observe; tools are never called and no code audit is performed.

### Tool changes

- 2026-10-13 `livonian_get_inflections` changed (inputSchema, outputSchema)
- 2026-10-13 `livonian_search` changed (outputSchema)

Full history: https://api.protogrid.dev/v1/servers/com.valksor%2Flivonian/changes

## Remotes

- https://livonian.valksor.com/mcp (streamable-http, auth none, reachable true, uptime30d 0.9745763)

## Packages

_none_

## Tools (4)

- `livonian_deromanize`: Find the real Livonian word(s) whose pronunciation-romanization matches this
clean-ASCII input — the inverse direction of `livonian_romanize`. Because
romanize is lossy, this is a LOOKUP against the words already in the dictionary
cache: attested spellings come back as source=dictionary. If none match, a single
best-effort reconstruction is returned (source=reconstruction, approximate=true) —
a guess, not attested. Coverage grows as words are cached, so an identical query
may return a reconstruction now and a dictionary hit later.

- `livonian_get_inflections`: Fetch the full declension (nouns) or conjugation (verbs) table for a word
identified by a search result's inflection handle (entry_id + word_class) — the
exact case, number, tense, not a guessed ending. Use this only when a search ran
without inline forms, or to fetch a match a broad search left un-expanded. Never a
first call — the handle values only come from livonian_search results, and a
default search already returns each match's table inline, so calling this
afterwards just re-fetches what you have.

Returns Markdown — a nominal gets one case × {singular, plural} table; a verb gets a table
per mood/tense (person × {affirmative, negative}, the negation auxiliary folded
into the negative) plus its principal-parts and participle lists — or a clean
not-found message. (Grammatical labels are normalised to English; Livonian forms
stay verbatim.) The same table rides along as structuredContent with the shape
{"result": <paradigm>} per the declared outputSchema — switch on
result.category ('nominal' | 'verbal' | 'not_found') before reading the body.
Results are cached server-side (repeats are instant) and
calls are rate limited — on a rate-limit error, wait a few seconds and retry.
Table content is from livonian.tech (CC BY-SA 4.0 — attribute if republished).

- `livonian_romanize`: Strip Livonian orthography down to clean, pronounceable ASCII for an
English-trained downstream (a voice/TTS, a search box). See the `text` parameter
doc for the exact letter mappings. Returns Markdown plus the romanized output as
structuredContent matching the declared outputSchema. Pure local transform: no
dictionary lookup, no network, and the output is always ASCII.

- `livonian_search`: Look a word up in the real Livonian–Estonian–Latvian dictionary and return only
attested content, so translations are grounded, not invented.

Search a meaning (in English/Latvian/Estonian) to find the Livonian headword, or a
Livonian word to confirm it exists and read its sense, part of speech and examples.
See the `query` and `search_language` parameter docs for how to phrase a query. By
default each match's full inflection table is returned inline, so one call usually
suffices; on a broad query only the first N tables expand (the rest are listed as
handles to fetch with livonian_get_inflections). Returns Markdown plus the same result as
structuredContent matching the declared outputSchema.

Results are cached server-side, so repeating a query is instant and free; a
first-time query reaches the live dictionary and calls are rate limited — on a
rate-limit error, wait a few seconds and retry instead of re-issuing immediately.
Dictionary content is from livonian.tech (CC BY-SA 4.0 — attribute if republished).