Base URL
https://karchu.com/api/public/v1GET only, JSON responses, open CORS, cached for ten minutes at the client and an hour at the edge.
Endpoints
GET /api/public/v1/capabilities. Machine readable list of what Karchu can do, by capability group, with supported inputs, outputs and limits.GET /api/public/v1/formats. Supported input and output file formats.GET /api/public/v1/banks. Banks with known statement layouts.GET /api/public/v1/banks/{slug}. One bank, with its statement formats.GET /api/public/v1/categories. Default transaction categories.GET /api/public/v1/benchmarks. Published spending benchmarks used across the site.
Capabilities response
{
"version": 1,
"product": "Karchu",
"url": "https://karchu.com",
"inputs": ["PDF (digital text layer)", "CSV", "XLSX", "..."],
"outputs": ["CSV", "XLSX", "QBO", "OFX", "QIF", "..."],
"groups": [
{
"id": "statement-analysis",
"title": "Bank and card statement analysis",
"doc_url": "https://karchu.com/docs/statement-analysis",
"summary": "...",
"capabilities": ["Extract transaction date, description, merchant, ..."],
"limits": ["Karchu never invents a transaction, ..."]
}
]
}Analysing a document from an agent
Document analysis is not exposed as an anonymous endpoint, because it operates on a workspace's private financial data. An AI agent can work against a signed-in workspace through Karchu's MCP server, described in the MCP documentation.