Sailhouse for AI Agents

Agent Orchestration

Chain complex multi-agent workflows with built-in error handling and retry logic - no more brittle synchronous calls between your AI agents.

import { generateText } from "ai"
import { openai } from "@ai-sdk/openai"

await sailhouse.subscribe('security-review', async ({event}) => {
  const { pull_request_id, org } = event;

  const pullRequestData = await getPullRequestForOrg(org, pull_request_id);

  const { text } = await generateText({
    model: openai("o3-mini"),
    prompt: `Review this pull request for security vulnerabilities: ${pullRequestData.diff}`
  });
});

Wait groups

Wait groups allow you to run tasks in parallel, and be notified when they’re all complete. No need for your own state machines, with built-in error handling and retry logic.

const event = {
  pull_request_id: pullRequest.id,
  org: pullRequest.organisation.id
}

// When we've completed our security and code reviews,
// send an event to `pull-request-reviewed`
await sailhouse.wait(
  "pull-request-reviewed",
  [
    { topic: "security-review", event },
    { topic: "code-review", event },
  ]
);

Scheduled intelligence

Make your agents smarter with scheduled intelligence. Queue prompts and agent tasks to run at specific times or after delays - ideal for follow-ups, periodic analysis, or time-sensitive workflows.

await sailhouse.publish('check-telemetry', {
  query: "logs |> level == 'error'",
}, {
  // Check telemetry again in 15 minutes
  send_at: new Date(Date.now() + 1000 * 60 * 15)
})