Build durable AI agents with Flue's programmable TypeScript harness. Write once, deploy anywhere, use any LLM.
'use agent';
import { useModel, usePersistentState, useSandbox, useSkill, useTool } from '@flue/runtime';
import { local } from '@flue/runtime/node';
import triage from '../skills/triage/SKILL.md';
// This is an agent.
export function IssueTriage() {
// Define your agent with hooks.
// Dynamically attach capabilities to your agent.
useModel('anthropic/claude-sonnet-5-0');
useSkill(triage);
// Store persistent data on each agent.
// Update your agent capabilities as state evolves.
const [isSandbox, setSandbox] = usePersistentState('isSandbox', false);
useTool({name: 'handoff', description: '...', run: () => setSandbox(true)});
if (isSandbox) useSandbox(local());
// Return your agent instructions. Flue handles the rest.
return `Triage the bug report end-to-end...`;
}Flue is powered by Pi, the trusted agent harness used by OpenClaw and millions worldwide. Pi gives you the tools you need to unlock real agent autonomy, moving beyond simple chatbots.
Pi is the core of our open tech stack including Durable Streams, Vite, and a flexible Sandbox API. Deploy Flue anywhere with confidence.
Agents can’t die when the server goes down. Flue records every session in a durable stream, then safely resumes interrupted work when the runtime comes back online. Run one agent or a multi-agent swarm on the same durable foundation.
Build agents that can safely take action, maintain continuity, and connect to the systems where work already happens.
Build agents that can keep context across conversations and events as they autonomously work toward a goal. Each one is durable, stateful, and addressable over HTTP.
Agents are just functions. Script them with flue run and the SDK — locally, in CI, or from your own backend — or orchestrate them with Cloudflare Workflows, Inngest, and more.
A secure environment where agents run commands, edit files, and do real work.
Write data on each agent and update its capabilities as state changes.
Define specialized roles and let your agent delegate work to the right expert.
Agents call APIs, query data, and make changes — using the code you define.
Package expertise that agents load whenever a task needs guidance.
Connect agents to thousands of tools through the open Model Context Protocol.
Work survives crashes, restarts, and deploys — sessions resume automatically.
Export session traces to OpenTelemetry, Braintrust, Sentry, or your own observer.
Connect agents to where work happens — Slack, Teams, Discord, GitHub, and more.
Learn how to build, run, and deploy production-ready agents with Flue.