Quickstart
Read live markets in 30 seconds. Create one on-chain in 5 minutes.
You don't need an account, an API key, or to install anything to start reading. You need a Solana wallet with ~0.05 SOL to start writing.
Read a live market — 30 seconds
That's the whole read surface. Returns JSON with market addresses, pool balances, target pools, voter counts, expiry, current resolution status, and links to the on-chain accounts. Drop the response into your agent's context and it can answer questions about live markets immediately.
For a single market by id:
No auth, no rate-limit-on-reads outside reasonable IP throttling (60/min), and the data is cached so the latency is low.
Vote on an existing market — 5 minutes
The voting instruction (buy_yes / buy_no) is permissionless on-chain. Any Solana keypair with ≥0.01 SOL can stake. The web UI wraps this in Privy auth, but you can call it directly.
This is the exact same code path the web app uses — the only difference is auth. Both go through the program's buy_yes instruction. No SDK is required.
Create a market — 5 minutes (founder path)
Creating a market is also permissionless. You pay 0.015 SOL to plant. The pattern is identical to voting, with two extra inputs (IPFS CID for metadata + target pool + duration).
The metadata you pin to IPFS should be a JSON object like:
The web app's projects-create endpoint handles the IPFS pin for you if you'd rather use it.
Crank an expired market — 30 seconds
If you see a market past its expiry timestamp that nobody has resolved yet, you can crank it forward. This is permissionless and the instruction takes no signer:
After expire succeeds, anyone can call resolve_market to determine the outcome. The pump.fun launch CPI inside resolve_market requires the founder pubkey as a non-signer account (see On-chain program for the full instruction reference).
Claim winnings — 30 seconds
If you voted on the winning side, your share is sitting in the program waiting for you to claim:
The full account list for each instruction is in the on-chain program reference.
Where to go next
- Reading more state than the markets endpoint exposes? See the Public read API for the full endpoint surface and JSON shape.
- Building an integration that needs the on-chain truth? See the On-chain program reference for every instruction, its account list, and its argument layout.
- Wiring this into an AI agent? See Agent integration for
llms.txt, the planned MCP server, and the recommended hand-off pattern (agent prepares, user signs).