CONAB - Companhia Nacional de Abastecimento¶

Overview¶

Field	Value
Institution	Ministério da Agricultura
Website	conab.gov.br
agrobr access	Direct (public XLSX)

Data Origin¶

Source¶

URL: https://www.gov.br/conab/pt-br/atuacao/informacoes-agropecuarias/safras/safra-de-graos/boletim-da-safra-de-graos
Format: XLSX (Excel spreadsheets)
Access: Public, no restrictions

Surveys¶

CONAB publishes monthly crop surveys:

Month	Survey
October	1st Survey
November	2nd Survey
December	3rd Survey
January	4th Survey
February	5th Survey
March	6th Survey
April	7th Survey
May	8th Survey
June	9th Survey
July	10th Survey
August	11th Survey
September	12th Survey

Available Data¶

Crops¶

Planted area (thousand hectares)
Harvested area (thousand hectares)
Yield (kg/ha)
Production (thousand tonnes)

Supply and Demand Balance¶

Opening stock
Production
Imports
Consumption
Exports
Closing stock

Usage¶

Crops by Product¶

import asyncio
from agrobr import conab

async def main():
    # Soybean crop data
    df = await conab.safras('soja')

    # Specific crop year
    df = await conab.safras('milho', safra='2025/26')

    # Filter by state
    df = await conab.safras('soja', uf='MT')

    # With metadata
    df, meta = await conab.safras('soja', return_meta=True)

asyncio.run(main())

Supply/Demand Balance¶

# Balance for all products
df = await conab.balanco()

# Balance for a specific product
df = await conab.balanco(produto='soja')

Brazil Totals¶

# National totals by product
df = await conab.brasil_total()

Schema - Crops¶

Column	Type	Description
`fonte`	str	"conab"
`produto`	str	Product name
`safra`	str	Crop year (e.g. "2024/25")
`uf`	str	State code
`area_plantada`	Decimal	Thousand hectares
`area_colhida`	Decimal	Thousand hectares
`produtividade`	Decimal	kg/ha
`producao`	Decimal	Thousand tonnes
`levantamento`	int	Survey number (1-12)
`data_publicacao`	date	Publication date

Available Products¶

produtos = await conab.produtos()
# ['soja', 'milho', 'arroz', 'feijao', 'algodao', 'trigo', ...]

Available States¶

ufs = await conab.ufs()
# ['AC', 'AL', 'AM', 'AP', 'BA', 'CE', 'DF', 'ES', 'GO', ...]

Available Surveys¶

levs = await conab.levantamentos()
for lev in levs[:5]:
    print(f"{lev['safra']} - survey #{lev['levantamento']}")

Production Cost¶

Production cost per hectare spreadsheets, published on gov.br/conab.

# Detailed soybean cost in MT
df = await conab.custo_producao("soja", uf="MT")

# Totals (COE, COT, CT)
totais = await conab.custo_producao_total("soja", uf="MT", safra="2024/25")

Schema - custo_producao¶

Column	Type	Description
`cultura`	str	Crop name
`uf`	str	State code
`safra`	str	Crop year (e.g. "2024/25")
`item`	str	Cost item
`categoria`	str	Category (insumos, operacoes, mao_de_obra, etc)
`valor_ha`	float	Value per hectare (R$/ha)
`unidade`	str	Item unit

Status (Mar/2026)¶

The grain spreadsheets (soja, milho, cafe, algodao) on gov.br are loaded via dynamic JavaScript and do not have .xlsx links accessible via scraping. Special crops (abacaxi, banana, cebola, cacau, tomate, etc.) work normally.

The parser (v2) supports 4 spreadsheet formats: - Standard format — header in 1 row (Item/Especificacao, Unidade, Valor Total/ha) - Format A — header split across 2 rows (DISCRIMINACAO + R$/ha in separate rows) - Format C — compact 1-row header (DISCRIMINACAO, CUSTO POR HA, CUSTO/kg) - Format D — header split across 3 rows (A PRECOS DE + DISCRIMINACAO + R$/ha). Detection via best-quality selection among 2-row candidates

Historical Series (v0.8.0)¶

Historical crop data since ~1976, published in Excel spreadsheets (.xls legacy). The parser automatically detects the format (OLE2/BIFF → xlrd, OOXML → openpyxl with calamine fallback).

# Soybean historical series
df = await conab.serie_historica("soja", inicio=2020, fim=2025)

# Filter by state
df = await conab.serie_historica("soja", inicio=2020, uf="MT")

Schema - serie_historica¶

Column	Type	Description
`safra`	str	Crop year (e.g. "2024/25")
`produto`	str	Product name
`uf`	str	State code
`area_plantada`	float	Thousand hectares
`producao`	float	Thousand tonnes
`produtividade`	float	kg/ha

Cache¶

Aspect	Value
TTL	24 hours
Max stale	30 days
Policy	Fixed TTL

Update¶

Aspect	Value
Frequency	Monthly
Publication	Generally between days 10-15

Datasets¶

serie_historica_safra — wraps conab.serie_historica() (32 crops)