Tidepool: a 1Click-compatible testnet sandbox for NEAR Intents

The problem

NEAR Intents and the 1Click API are great to build on, but there’s no testnet. Testing an integration (quotes, deposits, status polling, refunds) or an AI agent that moves money means using real mainnet funds. This keeps coming up in Discord.

What I built

Tidepool is a community-built, free testnet version of NEAR Intents, with a 1Click-compatible API and an Agent API. It’s live: www.usetidepool.tech

1Click-compatible API (/v0)

  • Same API as 1Click. /v0/tokens, /v0/quote, /v0/deposit/submit, /v0/status, /v0/generate-intent, /v0/submit-intent, /v0/orders, /v0/auth/* and /v0/account/*, with the same fields. The official 1Click TypeScript SDK works by changing one line (OpenAPI.BASE).
  • Every swap type. EXACT_INPUT, EXACT_OUTPUT, FLEX_INPUT and ANY_INPUT, plus app fees, MEMO deposits and LIMIT orders.
  • Real on-chain settlement. Swaps settle through a deployment of the official, unmodified Intents verifier (now defuse v0.4.5) on testnet. A maker bot fills quotes in seconds.
  • Failure paths like mainnet. Partial deposits, expired quotes and overpays are refunded correctly, with the matching statuses and refund reasons.

Agent API (/v1), matching NEAR Intents’ Agent API

  • Get a free key instantly with POST /v1/keys, no sign-up needed.
  • Policy-bound agent wallets with spend limits and allowed/blocked destinations, plus idempotency keys.
  • Swap, transfer, deposit and withdraw, with dry runs.
  • Simulated failures, so you can test how your agent handles a FAILED status.
  • Works with @near-intents-agent-api/sdk by changing the baseUrl.
  • Already used by builders to crash-test AI payment agents.

Cross-chain testnet bridges
Payouts and deposits on Base, Arbitrum and Ethereum Sepolia, Solana, Stellar, Sui, XRP Ledger, Bitcoin testnet4 and Zcash testnet.

Test-token faucet
USDC, USDT, wNEAR, wBTC, ETH and XAUT on NEAR, plus test USDC/USDT on the other chains. Agents can fund themselves with one call.

One-call demo
POST /v0/demo/run runs a real testnet swap and limit order and returns every request and response, with explorer links.

Docs AI agents can read
llms.txt, llms-full.txt, AGENTS.md, an OpenAPI spec, copy as Markdown, and open in ChatGPT/Claude: www.usetidepool.tech/docs

The workflow it enables

Build and test your full integration or agent on Tidepool for free. Then point the SDK at mainnet and do one final run with small amounts.

Tidepool is community-built and not affiliated with NEAR. I’d love feedback from builders and the Intents team: what’s missing, what breaks, and what you’d like to see next.