Flue

Flue gives you a composable TypeScript harness for autonomous agents, treating agents as declarative functions powered by hooks like useSandbox and useSkill. You get multi-file editing, durable state that resumes after crashes, and support for Model Context Protocol client tooling across Node.js, Cloudflare Workers, and headless CI. Because it relies on BYOK model routing rather than a turnkey bundled runtime, you take on full architectural responsibility for your provider bills and orchestration.

You will appreciate this if you want code-first control over sandboxes and agent lifecycles in JavaScript. Pick LangGraph or Claude Code directly if you want broader ecosystem integrations or an out-of-the-box coding agent instead of building your own framework.

Flue is useful for TypeScript developers who need stateful background agents. The framework abstracts model execution, MCP tool calls, and durable state into a declarative hook model. State persists across server restarts through recorded session streams. However, configuring execution requires caution. Running useSandbox(local()) grants filesystem access without container isolation. The architecture depends entirely on external API keys and brings no built-in terminal or browser.

Adopting Flue requires committing your workflows to its custom runtime abstractions. The open-source model keeps entry free, but unconstrained multi-agent delegation risks uncontrolled API token consumption during autonomous loops.

Flue targets TypeScript teams building autonomous harnesses. It provides a functional API modeled around explicit hooks—such as useModel, useSandbox, and usePersistentState—to declare execution environments and tools. Two architectural choices stand out: sessions persist into durable streams to allow recovery after unexpected restarts, and integrations are exposed through the Model Context Protocol. Documented runtime targets include Node.js, Cloudflare Workers, and headless continuous integration pipelines.

The trade-off is architectural opacity regarding model governance. While Flue abstracts tools, sandboxes, and skills, evaluation rigor is unaddressed; no standardized benchmarks or comparative harnesses are published in the documentation.

Flue is useful for TypeScript teams building autonomous workflows. Developed by Astro Technology Company, it packages sandboxes, tools, skills, and Model Context Protocol support into a functional composability pattern. Two deciding factors drive adoption: durable execution that resumes interrupted sessions over durable streams, and the portability to run agents locally, in CI/CD pipelines, or on hosted runtimes like Cloudflare Workers.

The trade-off sits squarely in the economics. Flue is open source and bring-your-own-model, which means zero software cost today but zero captive pricing power for the vendor. If monetization never materializes, you will bear long-term maintenance costs yourself. Position: use the framework freely, but keep agent boundaries clean in case corporate backing wanes.

Flue is useful for internal platform teams looking to assemble customized TypeScript agents rather than purchasing turnkey seats. It handles agent harness primitives like sessions, durable recovery, subagents, and Model Context Protocol client connections. Deployment targets include local CLI runs, Node.js, Cloudflare Workers, and headless CI pipelines, which fits existing deployment patterns without introducing proprietary seats.

The trade-off is operational ownership. The framework is open-source and free, but your team pays the unmetered model provider invoices under a bring-your-own-key model. Without native SSO or vendor audit logging out of the box, compliance rests entirely on your custom backend deployment. Approved with conditions.

Flue is useful for TypeScript developers building autonomous agents that need state persistence and external tools. You declare agents as plain functions using composable hooks like useModel, useSandbox, and useTool. Having an MCP client right in the harness means hooking up standard tool definitions without custom glue code. It also supports multi-file editing and runs in headless CI or Cloudflare Workers.

The trade-off is model freedom. While it supports bring-your-own-key for cloud backbones like Claude Sonnet, it lacks local model execution. If you demand fully offline runs on your own hardware, this harness will not deliver.