Guide: Autonomous Research Agent

A representative flow: a user deploys a research agent, funds it with TOKE, and lets it pay for data and compute without further approval.

The flow

1. A user deploys a research agent and funds it with a Tokelio budget.
2. The agent spends TOKE to access premium data feeds and compute for analysis.
3. On completion, the agent settles a small execution fee and returns results to the user.
4. Unused budget remains available for the agent's next task — no further human approval.

In code

// 1. Deploy + fund
const agent = await tokelio.agents.deploy({
  name: "market-research",
  budget: 300,
  policy: { maxPerTask: 30, dailyCap: 120 },
});

// 2. Agent pays for inputs
await tokelio.settle({ task: "data:premium-feed", amount: 8, to: "provider:quill" });
await tokelio.settle({ task: "inference:analyze", amount: 14, to: "provider:atlas" });

// 3. Settle the execution fee, return results
await tokelio.settle({ task: "exec:report", amount: 2, to: "agent:market-research" });

Why this matters

The agent operates like an employee with a corporate card and a spending policy: it transacts continuously, stays inside its limits, and never blocks on a human to approve a $0.08 data call.