Last updated:
SUZUNA is a Telegram-based AI wallet-intelligence service for understanding activity on the BNB Chain. Launched on September 9, 2026, it lets users associate a public wallet address with a personal AI agent and ask questions about the wallet in natural language. The product is designed to explain wallet activity, not to control wallets, execute transactions, or manage assets.[1][2][3]
Users deploy an agent through the SUZUNA Telegram bot by submitting a public BNB Chain address. The address identifies the wallet the agent represents. Deployment is an off-chain database operation and requires no wallet connection, signature, transaction, or gas fee. Users can then ask about activity over a particular period—for example, recent purchases, the largest transactions, or changes during the previous seven days—and follow up with questions about the same wallet.[2][1][4]
The service analyzes balances, incoming and outgoing transactions, token transfers, and contract interactions, including approvals, swaps, and claims. Rather than return isolated transaction records, SUZUNA says it identifies relevant signals and context across events and over time. An AI Analysis Layer then uses this information to produce a natural-language explanation. The underlying blockchain records remain the basis for that explanation.[3][5][6]
SUZUNA’s initial product focuses on BNB Chain, Telegram, and wallet intelligence. It retrieves relevant blockchain information when needed rather than maintaining a permanent copy of each wallet’s complete history. The project says figures in its answers are read from the chain and amounts are supplied in readable units, so the language model does not need to perform token-decimal arithmetic.[2][7][4]
The project describes its agents as read-only: they do not hold assets, access private keys, sign transactions, transfer funds, trade, copy trades, or provide portfolio management. Users need not prove ownership of a wallet because the service analyzes activity at a public address rather than accessing the wallet itself.[6][8][3]
In the repository’s description of the question-answering workflow, the AI model selects from five tools: retrieving the current native BNB balance, exact token balances, recent BNB transactions, token transfers, and contract interactions ranked by use. The wallet address is bound to the agent on the server, rather than passed to a tool as a model-selected argument. The project says this prevents the model from choosing a different wallet and that answers are intended to be traceable to tool-result rows.[4]
The repository identifies TypeScript, grammY, Fastify, Supabase, viem, Nodereal, DeepSeek, and Solidity among the project’s technologies. Wallet analysis uses BNB Chain mainnet data; usage-checkpoint publication is configured separately. The project’s website says its contract is deployed on BNB Chain mainnet, has public source code, and has no update path.[4][7]
The repository documents several technical constraints. The indexer query is capped at two million blocks, which the project estimates covers about 70 days of activity. Chain data is cached in process memory, so each service instance maintains its own cache. An unqualified question is routed to the user’s newest agent; the user cannot specify another agent in the question. Token balances rely on the indexer without a second-source cross-check, and tests cover the checkpoint engine but not the Telegram, agent, or chain layers.[4]
SUZUNA says it collects and retains only the information needed to operate agents and measure product use. An agent record associates an internal user with a wallet address and includes information such as the agent’s status and activity timestamps, but no wallet credentials. The service records lightweight events, including agent deployment, agent use, and AI queries. It says it does not store the text of questions, including in application logs.[8][4]
The project says it does not retain raw Telegram user IDs. When a message arrives, the ID is hashed using HMAC-SHA256 with a server-side secret, after which the raw ID is discarded. The stable hash lets the same Telegram account resolve to the same internal user. SUZUNA also says it does not request or store Telegram usernames, display names, or profile photos. The bot can reply to users but cannot initiate contact, send announcements, or notify them when a tracked wallet moves.[8][4]
On a seven-day cycle, SUZUNA calculates five usage measures: cumulative users who have started the bot, cumulative agents created, distinct weekly active users, distinct agents used during the week, and AI queries made during the week. The project publishes these metrics to a smart contract as append-only checkpoints. It says confirmed checkpoints cannot be updated or deleted, and a database trigger enforces the same rule in its database. Each checkpoint is intended to show which figures SUZUNA published at a particular time.[4][9]
The repository describes tests covering checkpoint period calculations, idempotency, duplicate submissions, and frozen metrics. The publication test harness has preflight, publish, and verify stages. It compares database data with contract storage, transaction receipts, and emitted event fields.[4]
SUZUNA announced that it was live on September 9, 2026. In the launch announcement, the company reported 820 total users, 410 weekly active users, 1,060 agents deployed, and 1,554 AI queries. It also said its first usage checkpoint had been recorded on BNB Chain.[3]
On September 12, 2026, the company said the product and its code were public and that development was taking place in the open. On September 25, it described a separation between off-chain usage processing and on-chain verification: a backend records events and calculates metrics, then publishes weekly checkpoints to BNB Chain as a verifiable usage history.[3]
On November 8, 2026, SUZUNA reported more than 13,000 agents deployed. It said a public dashboard tracked users, agents deployed, weekly active users, active agents, and AI queries. These adoption figures were reported by the company.[3]
The GitHub repository was created on August 29, 2026. Its README sets out contribution principles, including read-only agents, avoiding raw Telegram IDs and stored question text, deriving metrics from events, keeping confirmed checkpoints immutable, and making AI-generated figures traceable to tool results.[4]
SUZUNA’s documented initial revenue model centers on paid Pro subscriptions. The Free plan allows up to two agents. After the token generation event (TGE), the planned Pro plan is to allow up to six agents, with subscriptions paid in the project’s intended utility token, $SZN. The Pro price is to be announced before subscriptions become active.[10][9]
The documentation gives $SZN a fixed supply of 1 billion tokens. It describes the intended distribution as supporting community growth, product adoption, development, liquidity, and contributor alignment through controlled distribution and vesting schedules. The token is planned to provide access to Pro and serve as its subscription payment asset after the TGE.[9][10]
SUZUNA described a partnership with Tilted, which it characterized as infrastructure for generative game creation. Tilted’s flagship product, Vera, was described as an agent that builds games from spoken briefs inside Unreal Editor 6.[3]