agents.do
AI Agents That Execute Autonomously
Deploy intelligent agents that analyze data, make decisions, and execute tasks with natural language commands.
import { $, db, ai, on, send } from 'sdk.do' // Agent detects churn risk and takes actionon($.Customer.SignalDetected, async (signal) => { const customer = await db.get('customers', signal.customerId) const assessment = await ai.generate({ prompt: `Analyze customer health for ${customer.name}`, schema: $.CustomerSuccess.assesses.Health }) if (assessment.riskLevel === 'high') { await send($.CSM.assigns, { customerId: customer.id, priority: 'urgent' }) await send($.Email.sends, { to: customer.email, body: assessment.outreachMessage }) }})
One of 48 primitives
agents.do sits in Agents: reason, research and act. Every .do shares one grammar, Business-as-Code, and one SDK.
Data
Records, vocabulary and search
Insight
Measure, test and trace
Frequently asked questions
Agents make independent decisions, execute multi-step tasks, and adapt to changing conditions without constant oversight. Configure boundaries, require approval for high-stakes actions, and let agents handle routine operations 24/7.
Agents analyze data, make decisions, call APIs, manage workflows, and interact with users via chat or email. They understand natural language, interpret semantic patterns, and execute tasks across your entire tech stack.
Set explicit boundaries and require human approval for high-stakes actions like payments or deletions. Monitor all decisions in real-time with full audit logs, rollback capabilities, and dry-run mode.
Yes, agents communicate via semantic protocols and delegate tasks to specialists using $.Agent.delegate.Task patterns. Build multi-agent systems where each agent handles their domain expertise.
Choose from GPT, Claude, Gemini, or bring your own models via API. Agents automatically route tasks to the best model based on cost, latency, and capability.
