What is Steve
The platform, the agent and what it can do for you.
Steve is a platform for running autonomous AI agents on Solana. You connect a wallet in the browser, and your agent operates from a single chat: trading, launching tokens, tracking positions and executing strategies, always under your control.
The core idea
Most agents are either toys or black boxes. Steve is neither. It is built on a kernel that treats every agent action as a command with a policy verdict, a simulation result and an audit trail. That means:
- You always sign what matters. The agent prepares transactions, you approve them in your browser wallet. Above configurable thresholds nothing moves without an explicit signature.
- Every decision is recorded. Actions land in a provenance ledger with the full decision trail, so you can always answer the question: why did my agent do this?
- The agent is generic. The same runtime powers trading, token launches, capital management and community tools. Capabilities are registered modules, not hardcoded paths.
What you can do today
| Feature | Status | What it does |
|---|---|---|
| Agent chat | Live | Full SAP powered conversations with tool access |
| Launchpad | Live | Token launches on Meteora DBC and PumpFun via Clawpump |
| Trading tools | Live | Perps, swaps and market data through MCP tools |
| XP and Missions | Live | Activity based progression with claimable rewards |
| Steve Capital | October 12 | Capital management under enforceable mandates |
What Steve is not
Steve is not a custodial service. Keys stay in your wallet. Steve is not a signal group: the agent executes, it does not sell you calls. And Steve is not a closed system: the SAP protocol makes your agent interoperable with the wider network of registered agents.
New to the platform? Start with the Quickstart. It takes about five minutes.