From zero to a first answer in five minutes.
One API key, one command, and your agent is reading official sources from Argentina, Brazil, Chile and Mexico. This page is the whole reference: tools, parameters, errors and limits, with real responses from the server.
Quickstart
1. Get a key
Pick a plan on the plans page and sign in with your email: we send a link, there is no password. In the dashboard, create an API key. It is shown once, so store it as a secret. Every key starts with jn_live_.
2. Connect your client
The server speaks MCP over streamable HTTP and authenticates with the key in a header. Any client that accepts a remote HTTP server with headers works.
Endpoint, transport and header are all a client needs. MCP accepts X-API-Key: <key> or Authorization: Bearer <key>.
Claude on the web and desktop, and ChatGPT, add remote servers through OAuth, not API keys. OAuth is on its way; until then, use one of the clients above.
3. Your first REST call
The same key opens the REST API. One call, one JSON answer:
Try these with your agent
- “Find the Supreme Court tesis on suspension of the challenged act in amparo and give me the official link for each one.”
- “Which laws amended Argentina's consumer protection law, Ley 24.240? List them by date.”
- “Search SCBA rulings on dismissal during pregnancy since 2020 and summarise the criterion, citing each ruling.”
The four tools
Real responses from the service, trimmed for length.
search
Searches every source your plan covers. Returns ranked results, each with the official link.
| Parameter | Type | Default | What it does |
|---|---|---|---|
query | string | required | Keywords or a citation (“Ley 24.240”). |
country | string | — | ISO code: AR, BR, CL or MX. Without it, every country in your plan. |
source | string | — | A source id from list_sources, e.g. AR/SCBA. |
type | string | — | Document type as returned in results, e.g. legislation, case_law. |
date_from · date_to | string | — | ISO dates, YYYY-MM-DD. |
tipo_norma · numero_norma | string | — | Exact InfoLEG filters: Ley and 24240. |
limit | integer | 10 | Results per page, 1 to 50. |
offset | integer | 0 | Results to skip, up to 1,000. |
Each result carries id (use it with get_document), source, type, title, citation, date, issuer, url (the official source), a text_snippet with matches wrapped in <mark>, and a relevance score.
get_document
Fetches a document's metadata and text by the numeric id from search. Long texts come in pages.
| Parameter | Type | Default | What it does |
|---|---|---|---|
id | integer | required | The id from a search result. |
max_chars | integer | 20000 | Characters of text per call, up to 100,000. |
offset_chars | integer | 0 | Where to start reading. Pass the previous text_next_offset. |
text_length is the full length; text_next_offset is null once you have read to the end.
list_sources
Lists the sources your key can read, with totals per country and the date of each source's last update run.
| Parameter | Type | Default | What it does |
|---|---|---|---|
country | string | — | AR, BR, CL or MX. Without it, every country in your plan. |
doc_count changes every day, so the example hides it. last_run is when that source was last updated.
get_relations
Walks Argentine legislation: which norms amend a law, and which laws it amends. Data from InfoLEG.
| Parameter | Type | Default | What it does |
|---|---|---|---|
id | integer | required | The id of an InfoLEG document (source = AR/INFOLEG-SEED). |
direction | string | both | modificada_por (amended by), modifica_a (amends) or both. |
limit | integer | 50 | Relations per page, 1 to 200. |
cursor | string | — | The previous next_cursor. |
Each relation points to the other norm by related_document_id, which you can pass to get_document.
REST API
The REST API reads the same data as the tools. Send the key in X-API-Key: REST does not accept Authorization: Bearer.
Base URL: https://api.jurisnet.ai/api/raw
| Method and path | Tool | Notes |
|---|---|---|
GET /search?q=… | search | Same filters, with q instead of query. limit 1–100, default 20. No offset: for deep pagination use MCP. |
GET /documents/{id} | get_document | All the available text in one response, plus raw_payload (the source's own metadata), fetched_at and last_seen_at. |
GET /sources | list_sources | Includes each source's status. |
GET /documents/{id}/relations | get_relations | direction, limit (1–200) and cursor. |
Search and filters
- Write short keyword queries in the language of the documents: Spanish for Argentina, Chile and Mexico, Portuguese for Brazil.
- Pass
countrywhenever you know it, andsourcewhen you know where the answer lives. It narrows the search and makes it faster. - Cite norms by number:
Ley 24.240,Decreto 70/2023. - Responses carry
total, the number of matches. With enhanced search, responses carryhas_moreinstead, andmatch: "any_terms"when no document contained every word and the results fall back to documents with some of them.
Pagination
search: raiseoffsetbylimiton each call. Up to 1,000 results deep.get_document: call again withoffset_chars=text_next_offsetuntil it isnull.get_relations: passnext_cursorascursoruntil it isnull.
Errors and limits
Authentication and limit errors come back as HTTP status codes:
Inside a tool call, problems with the request come back as a normal result with an error field: a document that does not exist or is outside your plan, a country your plan does not include, an invalid cursor. Your agent can read the message and correct itself.
The REST API uses HTTP status codes for everything:
Limits
- Solo: 30 requests per minute per key, 1 active key.
- Builder: 120 requests per minute per key, up to 5 active keys.
- Over the limit you get
429withRATE_LIMITED. Wait a few seconds and retry. - A revoked key answers
401. Keys are revoked when the subscription ends. - Each key reads the countries in its plan: up to 2 of your choice with Solo; with Builder, up to 5, which today means all four available (AR, BR, CL, MX). Asking for a country outside your plan returns an error.
Full text, summary or record
Every document carries its metadata, citation and official link. How much text comes with it depends on the source: in Brazil most documents include the full text; in Argentina, Chile and Mexico most are records, with the text where the source publishes it.
- Full text:
textholds the whole document, as published by the source. - Partial:
textholds a summary written by the source (a sumario or a tesis), or the text for only part of the collection. - Record:
textholds the metadata and a short description. The full document is aturl, often an official PDF.
How to cite
- Cite the
url: it is the official source. JurisNet is the road to the document, not the document. - Build the reference from
titleorcitation,issueranddate. - Before relying on a norm, check on the official site that it is still in force and that no later text replaced it.
JurisNet returns documents from official sources. It does not give legal advice, and it does not guarantee that a document is complete, current or in force.
Something missing or wrong here? Write to us · help@jurisnet.ai