Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.shuriken.trade/llms.txt

Use this file to discover all available pages before exploring further.

Shuriken Labs is the no-code way to build a trading agent. Write a task prompt, pick the tools it can use, set the limits, and decide when it fires. No code, no server. You can be running in a few minutes.
Every Terminal user as of 2026-05-04 was given $5 USDC of AI billing credit. It’s already in your account. That’s enough to run most agents for a couple of weeks.

​
What an agent is

Task prompt

The instructions you write once. “Find tokens called by 3+ alpha groups in the last hour with under $5M cap. Show me the top 5.”

Tools

The same trading, lookup, and alpha tools you use yourself. Buys, sells, balances, token stats, group calls, on-chain data.

Triggers

When the agent fires. On a cadence, after your own trades, or via an API call.

Guardrails

The agent runs under a scoped agent key. Per-trade caps, daily limits, and wallet restrictions are enforced server-side.

​
The build, trigger, run loop

  1. Build the agent in Shuriken Labs. Pick a starter or write the task prompt yourself.
  2. Wire a trigger so the agent knows when to fire. See Triggers for the three options.
  3. It runs under your scoped agent key. Every limit and slippage cap you’ve already set still applies.
  4. You review every session in the run history, with the full transcript and tool calls.
The agent can never override the limits on its key. If a trade would exceed your per-trade cap, Shuriken rejects it before it executes. No matter what the prompt asks for.

​
Three ways to fire it

On a schedule runs the agent on a cron-style cadence. “Every morning at 09:00 UTC, summarize calls from my Discords overnight, filtered to under $1M market cap.” When I make a trade fires the agent right after you buy or sell. “After every buy on Solana, watch the position and tell me when it hits +50% or -25%.” Allow API calls lets you invoke the agent from a webhook, an MCP client, or any HTTP request. “Call the agent from a Telegram bot when it sees a contract paste, and have it run a rug check.” You can attach more than one. A research agent can be on a daily schedule and callable on demand from your Claude Desktop session at the same time.

​
Stop button and monitoring

The chat panel has a Stop button. Clicking it halts the upstream run, not just your local stream, so your billing stops at the same moment. The button shows a “Stopping…” loader for a second while the upstream confirms. Every run shows up in the agent’s session list with the full transcript, every tool call, and the result. Open a session to see exactly what the agent did and how long it took.

​
Credits and billing

Every Terminal user as of 2026-05-04 received $5 USDC of AI billing credit, retroactively. It’s already in your account. Top up at any time with USDC, or swap from the chain’s native gas asset (SOL, ETH, BNB) into USDC in one step from the top-up modal. Each run charges only the model time it actually used. The Stop button is the safest way to bound a run that’s drifting. If you hit zero balance, scheduled and trade triggers stop firing until you top up.

​
Get started

Build an agent

Open Labs and pick a starter, or build your own from scratch. Takes about 2 minutes.

Configure triggers

Wire your agent to a schedule, your own trades, or an API call.
New to agent keys? Create one first. Labs uses the same key system as the rest of Agent Kit, so the limits you set on the key apply to every agent that runs under it.