AI agent tokens and bot infra surge
Several agent-focused launches and tokens are being discussed on Solana: $ZEREBRO is positioning as an autonomous content‑farming agent, and $CLAWD aims to port Anthropic's Claude framework on‑chain with multi-tool integrations. The ecosystem recap also highlighted agentic projects like PenguBot for Telegram trading, dflow integrations for single-prompt trading apps, and MiloOnChains agents executing trades. (x.com) (x.com) (x.com)
On Solana, “AI agent” now means software that can call tools and move money, not just chat — and a cluster of new tokens and bot projects is being marketed around that shift. (github.com 1) (github.com 2) One layer is infrastructure for builders. SendAI’s open-source Solana Agent Kit says any model can perform “60+ Solana actions,” including token trades, lending, staking, bridges, airdrops and token launches. (github.com) Another layer is agent software that plugs those actions into large language models. The solana-clawd repository describes itself as a Solana rebuild of “Clawd Code,” runs as a Model Context Protocol server, and says clients can access 31 live Solana tools from apps such as Cursor, Visual Studio Code and Windsurf. (github.com) (anthropic.com) That is the setup behind tokens such as ZEREBRO and CLAWD. CoinGecko describes Zerebro as an autonomous system for creating, distributing and analyzing content across decentralized and social platforms, with a Solana token supply capped at 1 billion. (coingecko.com) Trading is the clearest consumer use case so far. PenguBot says its Telegram bot lets users place buys and sells with natural language, while dflow says its trading application programming interface powers “tens of billions” of dollars in transacted value on Solana. (pengubot.ai) (dflow.net) Projects are also starting to market agent-to-agent trading, where one bot hands portfolio decisions to another bot. The &milo “agent-to-agent portfolio manager” repository says outside agents can plug into its Solana-native trading system, which lists Jupiter and dflow in its execution layer and claims 100,000-plus transactions in the last month. (github.com) The pitch is speed and simpler interfaces. Instead of opening a wallet, a decentralized exchange and an analytics site, the user types a sentence, the model maps that request to on-chain actions, and the software executes or prepares the transaction. (github.com 1) (github.com 2) The harder part is verification. Several of the newest agent projects are being promoted first through GitHub pages, token listings and social posts, and many usage figures — including trader counts, assets under management and transaction volume — come from the projects themselves rather than from independently audited disclosures. (github.com 1) (github.com 2) For now, the Solana agent trade looks less like one product launch than a stack forming in public: toolkits for developers, execution rails for trades, chat interfaces in Telegram and tokens wrapped around the software. (github.com) (dflow.net)
