Explore API & gRPC

Two surfaces. One reaches the ledger, one reaches the book.

Both are gRPC, so there is no transport to choose. The only question is whether you need Canton itself — transfers, DvP, recurring payments — or Silvana Book's own execution, pricing and settlement services.

{ your.code }gRPCcantonagentic-apisilvana bookbook grpctransferorder
Which surface

Both are gRPC. They reach different things.

This is the one question worth answering before you write a line, and it is not a question about transports.

agentic-api

Reaches Canton itself

Programmatic access to the Canton Network through the Silvana Ledger gRPC service. Transfers, delivery-versus-payment, CIP-56 tokens, preapprovals, recurring payments, onboarding, atomic multicall, a cross-network bridge.

Use it when you are connecting a system that already exists. Any language with a gRPC client can call it, and you own the signing and the retry logic the SDK would otherwise have handled for you.

// prepare → sign locally → execute. The key never moves.

book-services

Reaches Silvana Book

Five services covering the venue rather than the ledger: execution, pricing, settlement, the ledger gateway and news. These are what an agent is actually talking to while it works an order.

Use them when the thing you need is the book — a price, a fill, a settlement state. The SDK is generated against these definitions, so building on it means using them without writing the client.

// execution · pricing · settlement · ledger-gateway · news

If you are building an agent rather than connecting a system, neither of these is your starting point — the SDK carries both for you.

Why through Silvana

One party to authenticate against.

Reaching Canton directly means standing up your own ledger client, your own signing, your own settlement watcher — and repeating the exercise per venue.

the-short-version.txt
  • One authenticated party reaching both the ledger and the book
  • Ed25519 message-level signing, with the key staying on your side
  • Reflection, so your tooling can discover the surface itself
  • Atomic multicall, instead of orchestrating a batch yourself
  • Two-phase execution, so a signature is never a blind one
  • The same definitions the SDK and the catalogue agents are built on
Next

Then put an agent on top of it.

The Agentic Kit is the templates, the runtime and the model that arrives trained — the layer that turns a surface you can call into something that runs without you.