BNBChain Studio¶

Scaffold, run, and deploy a two-layer blockchain seller on BNB Chain. BNBChain Studio (bnbagent-studio) lets you describe what you want in Claude Code or Cursor; the studio emits a working agent project that you own, then helps you develop, debug, and deploy it.

A seller agent earns on-chain by offering services over ERC-8004 (identity) + ERC-8183 (commerce) + x402 (payments), built on the BNBAgent SDK.

v0.0.1 is seller-only. The CLI (bag), runtime library (bnbagent_studio_core), read-only MCP server, and IDE skills ship today. Buyer product flows and a hosted console are deferred to v2.

⚠️ This project is under active development and may introduce breaking changes. It manages wallet keys and on-chain funds — start on testnet and use at your own risk.

What you get¶

Artifact	Package	Purpose
`bag` CLI	`bnbagent-studio`	Scaffold projects, manage wallets, run locally, deploy
Runtime library	`bnbagent-studio-core`	Wallet, ERC-8004/8183, x402, signing policy — imported by emitted agent code
MCP server	`bnbagent-studio`	15 read-only chain tools for your IDE
Skills	bundled in CLI	10 procedure playbooks for Claude Code / Cursor

Install one command:

pip install bnbagent-studio

pip auto-pulls bnbagent-studio-core. The CLI is also available via uv tool install bnbagent-studio.

Two-layer deploy model¶

AWS Bedrock AgentCore is invoke-only (no public HTTP routes, no background poll loop), so v1 splits the seller into two deployable artifacts:

            buyer                                   BNB Chain
              │ POST /apex/negotiate                   ▲
              ▼                                        │ funds / settles jobs
┌──────────────────────────────┐   InvokeAgentRuntime  │
│  Layer B — ERC-8183 Service  │ ──────────────────►  ┌┴─────────────────────────────┐
│  app/service/ → EC2/Fargate  │                      │  Layer A — the Agent         │
│  public · long-running       │ ◄────────────────── │  app/agent/ → AgentCore      │
│  KEYLESS · no LLM · no sign  │   signed offer /     │  non-public · invoke-only    │
│  /negotiate + funded-job     │   deliverable        │  the LLM + the SOLE signer   │
│  poller                      │                      │  quote / fulfill / settle    │
└──────────────────────────────┘                      └──────────────────────────────┘

Layer A — the Agent (app/agent/ → AWS Bedrock AgentCore): the LLM, memory, tools, and sole key-holder/signer. Invoked on action envelopes (quote / fulfill / settle). All signing is fixed entrypoint code — never an LLM-callable tool.
Layer B — the ERC-8183 Service (app/service/ → EC2/Fargate): a public, long-running, keyless container — /negotiate ingress, funded-job poller, and InvokeAgentRuntime client. Holds no key, runs no LLM, never signs.

Relationship to BNBAgent SDK¶

Layer	Package	Role
Protocol	bnbagent (BNBAgent SDK)	ERC-8004, ERC-8183, wallet ABC — pure protocol clients
Studio core	`bnbagent_studio_core`	Config, workflows, signing policy, audit log
Studio surface	`bag` CLI + MCP + skills + recipes	Scaffolding, ops, IDE integration
Your code	`app/agent/`, `app/service/`	Emitted by recipes; you own and edit freely

Use BNBAgent SDK directly when you want full control over protocol integration. Use BNBChain Studio when you want scaffolding, IDE skills, a two-layer deploy path, and safety defaults out of the box.

Prerequisites¶

Requirement	Why
Python ≥ 3.10	CLI and runtime library
Claude Code or Cursor	Studio is driven from your AI tool via skills
Node ≥ 20 + `npm i -g @aws/agentcore`	`bag init` / `bag dev` / `bag deploy agent` shell out to the native AgentCore CLI
A wallet password	`bag init` creates a local encrypted keystore; password lives in `WALLET_PASSWORD` env only

Optional: testnet funds (only for paid LLM models or on-chain settle — default auto/free Pieverse model needs no funds).

Documentation¶

Guide	Description
Quickstart	Install, scaffold, run locally, first negotiate
Demo	End-to-end weather-forecast seller walkthrough
Architecture	Six-layer stack, recipes, workspace layout
Configuration	`studio.toml`, `.env.local`, cross-layer sync
CLI reference	`bag` command groups and key flags
Deployment	Layer A (AgentCore) + Layer B (EC2) deploy path
Security	Keystore posture, signing policy, MCP read-only guarantee
Troubleshooting	`bag doctor`, common errors

Repository¶

https://github.com/bnb-chain/bnbagent-studio

Package¶

pip install bnbagent-studio

PyPI — bnbagent-studio · PyPI — bnbagent-studio-core

← Developer Kit overview