Plug real-time Solana memecoin signals into any AI agent. Pay per call in USDC with no account (x402), or connect over MCP, REST API, or webhooks. Our engine runs rug and safety checks before it calls a coin, and our track record is free to read.
Built for autonomous agents
Pique Signal scores new Solana tokens the moment they move with a convergence engine, and runs rug and safety checks before it calls a coin. It is built to be read by software, not just people: an agent can pay, read the board, and act without a human in the loop.
What you get
Why it matters to an agent
No human in the loop
The x402 per-call feed pays and reads in one flow — no login, no dashboard. A pass is bought once, then sent with each read.
Pay-per-call in USDC
No subscription. $0.02 for one read of the live board, or a pass ($0.50 for an hour, $5 for a day) for repeated reads plus Early Edge. The facilitator pays gas, so your wallet pays no SOL.
Early Edge
Our Early Edge list, included with a pass ($0.50 hour / $5 day) and with the $49 Gated and $99 Builder keys. Not in the $0.02 single read.
Rug & safety filters
We run rug and safety checks before we call a coin. Risky coins are filtered out or carry a warning tag. No check catches every rug.
Public track record
Free live JSON of the coins we flagged and how far they ran, from our first flag to the peak (/api/track_record_public) — verifiable, not vibes. Builder keys also get the backtest endpoints.
Real-time push
HMAC-signed webhooks and the Builder SSE stream come with the $99 Builder key. They deliver new signals the moment they fire.
Four ways to connect
Method
Best for
How
x402 Agent Feed
Autonomous agents, no account
Pay $0.02 in USDC per read, or buy a USDC pass, and read the live board. x402 docs →
MCP Server
Claude & any MCP client
Needs a paid key ($49 Gated or $99 Builder, from /unlock). Hosted (claude mcp add) or self-install; both include paper (practice) trading. On hosted, practice positions last only for that session and are lost when it ends or the server restarts. MCP docs →
Grok connector
Grok (Sign in with X)
Add https://grok.piquesignal.xyz/mcp in Grok — blue-check = 7-day gated trial, else the free board. Grok docs →
Builder API
Custom bots & backends
Needs a $99 Builder key. Builder REST feed, settled-results backtest, SSE stream, signed webhooks. API docs →
Any language works via the REST API and webhooks; MCP clients get a conversational interface to the same engine.
Get started
Fastest path for an autonomous agent — no account, no key. For one read, your agent requests the feed, pays the $0.02 x402 challenge in USDC, and sends the same request again with the payment:
GET https://piquesignal.xyz/api/x402/feed?view=gated
# -> returns an x402 402 challenge; pay $0.02 in USDC, then repeat the request to get the board
For repeated reads plus Early Edge, buy a pass once and read with it:
GET https://piquesignal.xyz/api/x402/pass?tenor=hour
# -> returns an x402 402 challenge; pay in USDC, then read:
GET https://piquesignal.xyz/api/x402/agent-feed?view=gated&limit=50
Authorization: Bearer pqx_...
Using Claude or a hosted MCP client instead? Add the server:
claude mcp add --transport http piquesignal https://piquesignal.xyz/mcp
Full endpoints, fields, prices, and examples are in the API docs.
For autonomous agents: a machine-readable menu of everything you can use or buy here is at /.well-known/agents.json.
FAQ
Can an AI agent use PiqueSignal without an account?
Yes. An agent can pay $0.02 in USDC for one read of the live board, or buy a pass ($0.50 for one hour, $5 for one day). No account, no API key, no human login.
How does an agent pay?
Via x402, an open pay-per-call standard. Two ways: (1) GET /api/x402/feed — you get a 402 challenge; send the same request again with the $0.02 USDC payment and it returns the board. (2) GET /api/x402/pass?tenor=hour|day — pay once and get a pass (shown only once, so save it), then read /api/x402/agent-feed with Authorization: Bearer <pass> until it expires. The facilitator pays gas, so your wallet pays no SOL.
Which agents and frameworks can connect?
Any MCP client (Claude Code and Claude Desktop, plus other MCP clients), and any language or backend via the REST API and webhooks.
Are the signals safety-checked?
Yes. We run rug and safety checks before we call a coin. Risky coins are filtered out or carry a warning tag. No check catches every rug.
Can I get signals pushed in real time?
Yes — a Server-Sent Events live stream, and HMAC-signed webhooks that POST each new signal to your HTTPS endpoint as it fires.
Is there a track record?
Yes. A free, live track record of the coins we flagged and how far they ran, measured from our first flag to the peak. Agents read it as JSON at /api/track_record_public (no key); people see it at /track-record. Builder keys also get settled outcomes through the backtest endpoints.