Your first screen

Overview

Four calls: find out what exists, resolve an identifier, pull a series, then put two different things on one axis. Everything here runs anonymously with the shared testing key.

export SILO_URL="https://zcjbtpxuhdekpwcxmepn.supabase.co/rest/v1"
export SILO_ANON_KEY="sb_publishable__yfFQsykAglrvc9GS6_PYw_B24ex437"

That key is shared — it identifies the project, not you, and goes on apikey only. See Quickstart.

1. Find out what is actually there

curl -s -X POST "$SILO_URL/rpc/coverage" \
  -H "apikey: $SILO_ANON_KEY" -H "Content-Type: application/json" -d '{}'

Read as_of (newest elapsed period landed) and complete_through (newest complete period, what default windows serve). A complete_through trailing today by a month or two for a fund dataset is CVM’s filing calendar, not an outage. Market coverage explains all four date columns.

2. Resolve an identifier

curl -s -X POST "$SILO_URL/rpc/lookup" \
  -H "apikey: $SILO_ANON_KEY" -H "Content-Type: application/json" \
  -d '{"p_query": "PETR4"}'

You get id, id_type, asset_class, name, isin, cnpj and any B3 tickers — the last from CVM’s published FCA map, never a guessed ticker↔CNPJ join. Note: an unknown identifier returns an empty array, not an error and never a substitute. If lookup gives you nothing, stop there rather than assuming the next call will work.

3. Pull a series

curl -s -X POST "$SILO_URL/rpc/quote_history" \
  -H "apikey: $SILO_ANON_KEY" -H "Content-Type: application/json" \
  -d '{"p_ticker": "PETR4", "p_from": "2025-01-01", "p_to": "2025-12-31"}'

import os, requests

URL = "https://zcjbtpxuhdekpwcxmepn.supabase.co/rest/v1"
H = {"apikey": os.environ["SILO_ANON_KEY"]}

rows = requests.post(f"{URL}/rpc/quote_history", headers=H, json={
    "p_ticker": "PETR4", "p_from": "2025-01-01", "p_to": "2025-12-31",
}).json()
print(len(rows), rows[0]["trade_date"], rows[-1]["close"])

import { createClient } from '@supabase/supabase-js'

const supabase = createClient(
  'https://zcjbtpxuhdekpwcxmepn.supabase.co',
  process.env.SILO_ANON_KEY,
  { db: { schema: 'api' } },
)

const { data } = await supabase.rpc('quote_history', {
  p_ticker: 'PETR4', p_from: '2025-01-01', p_to: '2025-12-31',
})

One year of daily prices is ~250 rows, comfortably inside the 1,000-row page. Ask for five years and the call refuses with 22023 rather than handing you the oldest 1,000 — page it with p_after. See the pagination guide before you write a loop. Note: prices are unadjusted (adjusted: false on every row). A return computed across a split date will be wrong unless you handle it yourself.

4. Put two different things on one axis

panel returns a long (id, date, metric, value) frame, so a listed company’s price and a fund’s NAV line up without a join.

curl -s -X POST "$SILO_URL/rpc/panel" \
  -H "apikey: $SILO_ANON_KEY" -H "Content-Type: application/json" \
  -d '{
        "p_ids": ["PETR4", "VALE3"],
        "p_metrics": ["close"],
        "p_from": "2025-01-01",
        "p_freq": "month"
      }'

Mix identifier kinds freely — tickers, option and termo codnegs, and fund CNPJs in one p_ids array. The fourteen metrics are close, volume, close_unit, close_return, nav, quota, delinquency, yield, inflows, redemptions, quotaholders, receivables, sacado_top1 and sacado_top25. Note: anonymously panel takes at most 3 ids per call — a fourth raises 22023 rather than quietly dropping one. Signing in raises it to 50.

Reading a null correctly

On this API a null is information, with three distinct meanings:

You see	It means
`null` outside a family’s applicable columns	not applicable — set by construction, e.g. `quota` for a FIDC
`null` inside them	a blank in that month’s filing
an empty array	no rows in that window, or an unknown id

None is ever filled with a guess. No forward-fill, no last known close, no interpolation. catalog().applicability lists which columns each family files; metric_coverage() gives the filed span per metric.

Paging without losing rows

Two cursors, one silent ceiling, and how to tell which cut you.

Reading a fundamental

A company’s financial statements end to end.

Market coverage

How far back each domain goes — and the one that is a ratchet.

For agents

The machine-readable contract, for tool-using models.

Overview

Data

Get started

Guides

API reference

Endpoints

Your first screen

Overview

1. Find out what is actually there

2. Resolve an identifier

3. Pull a series

4. Put two different things on one axis

Reading a null correctly

Next

Paging without losing rows

Reading a fundamental

Market coverage

For agents

​Overview

​1. Find out what is actually there

​2. Resolve an identifier

​3. Pull a series

​4. Put two different things on one axis

​Reading a null correctly

​Next

Paging without losing rows

Reading a fundamental

Market coverage

For agents

Overview

1. Find out what is actually there

2. Resolve an identifier

3. Pull a series

4. Put two different things on one axis

Reading a null correctly

Next