# ForestHub AI

> ForestHub is the Edge AI and Agents Orchestration Platform. Workflows are graphs and the LLM is one node among many, inspectable, replayable, auditable, bounded by design. The engine ships as a Go binary in a Docker image and runs on any Linux edge device.

## What is an edge agent?

An edge agent is an AI agent that runs where the work happens, on a Linux edge device, not in the cloud. It perceives local signals (sensors, MQTT/Modbus/OPC-UA), reasons over them, and acts on the physical system without a round-trip to a datacenter. ForestHub makes edge agents you can run in production: the workflow is a deterministic graph, the LLM is one node among many, inspectable, replayable, auditable, bounded by design. Where a cloud agent lets the model drive and hope, an edge agent runs a bounded graph where the LLM is a single, replaceable node and every action is verifiable. It is the only kind of agent you can safely put next to a machine. Full definition: [/en/edge-agents](https://www.foresthub.ai/en/edge-agents).

## TL;DR

ForestHub is a German deep-tech startup (founded 2025, Villingen-Schwenningen) building a graph-first orchestration platform for industrial AI agents. Author workflows as graphs in a visual builder, then deploy the engine as a distroless Docker image (linux/amd64 + arm64) on Linux edge devices such as gateways, NUCs, NVIDIA Jetsons, Raspberry Pis, and industrial PCs. Multi-LLM routing (Backend-Key / Bring-Your-Own-Key / On-Device). Industrial protocols (MQTT, Modbus TCP/RTU, OPC-UA) first-party. RAG via Knowledge Bases. GDPR-compliant. EU AI Act and Cyber Resilience Act readiness as architecture properties. **Open source, backend-optional**: Engine and LLM-Proxy ship as public packages with cross-language contracts. Self-host the stack on your own hardware, or use the managed SaaS for logging, device health, OTA deploys, and token accounting.

## What makes ForestHub different?

- **Open source, standalone, offline.** The `edge-agents` repository ([github.com/ForestHubAI/edge-agents](https://github.com/ForestHubAI/edge-agents), AGPL-3.0 + commercial dual-license) contains the Go engine, LLM-proxy, language-neutral contract, and the TS builder + `fh-builder` CLI. Runs without external services or accounts. The closed-source `fh-backend` is optional managed convenience on top.
- **Graph-first, not loop-first.** Loop-first agent frameworks (LangChain, OpenAI Agents SDK, Claude Code) make the LLM's loop the program. ForestHub inverts it: the graph is the program, the LLM is a node. Every possible decision path is a wire on the canvas; every run is replayable; every actuation is bounded by what was wired.
- **Engine, not codegen.** The engine interprets the workflow at runtime, with no per-device source-code generation and no recompile per target.
- **Linux-edge native by architecture.** Distroless Docker image, ~10 to 15 MB, multi-arch. Linux covers 95%+ of the edge/IoT market by volume. The same engine runs on Yocto, Buildroot, OpenWrt, Balena, and standard distributions.
- **Strict superset of loop-first.** A loop-first agent (one big prompt, many tools, run-to-completion) is available as a single agent node inside the graph. Constraint is opt-in.

## Key Questions & Answers

**Q: What are edge agents?**
A: An edge agent is an AI agent that runs where the work happens, on a Linux edge device, not in the cloud. It perceives local signals (sensors, MQTT/Modbus/OPC-UA), reasons over them, and acts on the physical system without a round-trip to a datacenter. ForestHub makes edge agents you can run in production: the workflow is a deterministic graph, the LLM is one node among many, inspectable, replayable, auditable, bounded by design. See [/en/edge-agents](https://www.foresthub.ai/en/edge-agents).

**Q: What is ForestHub?**
A: The Edge AI and Agents Orchestration Platform. Visual builder for graph-based workflows + Go-based engine in Docker that runs them on Linux edge devices.

**Q: Who uses ForestHub?**
A: Engineering teams at machine manufacturers (OEMs), system integrators, and industrial technology companies who need to ship LLM-using agents that pass an audit.

**Q: On what hardware does the engine run?**
A: Linux edge devices, `linux/amd64` and `linux/arm64`. Gateways, NUCs, NVIDIA Jetsons, Raspberry Pis, industrial PCs. One image, any Linux.

**Q: How is LLM behavior bounded?**
A: Agent nodes only see the tools the builder wires in. No generic `read_file`, no `exec`, no MCP server loading arbitrary capabilities. Deny-by-default at the architecture level, not enforced by sandboxes after the fact.

**Q: What LLM providers are supported?**
A: Three modes per workflow: Backend-Key (FH-managed proxy across OpenAI, Anthropic, Gemini, Mistral), Bring-Your-Own-Key (you pay the provider directly), or On-Device (a self-hosted inference endpoint in your network such as llama-server, vLLM, or Ollama). Different providers per node in the same graph.

**Q: How do I deploy a workflow?**
A: Three paths: API push from the builder (`POST /deploy`), Direct-IP within your LAN (`ENGINE_PUBLIC_URL`), or download a bootstrap bundle (`docker-compose.yml` + `manifest.json` + `README.md`) for offline / air-gapped install.

**Q: How does ForestHub differ from LangChain or OpenAI Agents SDK?**
A: Loop-first frameworks make the LLM's loop the program. ForestHub inverts that: the graph is the program, the LLM is a node. Inspectable, replayable, auditable, bounded by design.

**Q: How does ForestHub handle compliance?**
A: GDPR-compliant by architecture (data stays on-device unless you wire it elsewhere); EU AI Act readiness through visual auditability and structured event logs; Cyber Resilience Act readiness through Cosign-signed images, SBOM generation, and Trivy gate in the release pipeline.

**Q: Where is ForestHub based?**
A: Villingen-Schwenningen, Baden-Wuerttemberg, Germany. Registered at Amtsgericht Freiburg, HRB 734331.

**Q: Is ForestHub open source?**
A: Yes. The public repository is **`edge-agents`** at [github.com/ForestHubAI/edge-agents](https://github.com/ForestHubAI/edge-agents). It contains the Go engine + LLM-proxy (`go/`), the language-neutral OpenAPI 3.0.3 contract (`contract/`), the TypeScript workspace with the visual builder and `fh-builder` CLI (`ts/`), and a Claude Code skill that wraps the workflow CLI (`skills/`). The platform runs **standalone and offline, no external services or accounts required**. The closed-source `fh-backend` (governance, hosting, multi-tenant control plane) consumes this repo as managed convenience.

**Q: What license?**
A: `edge-agents` is dual-licensed. The public release is **AGPL-3.0-only**, including the AGPL's network-copyleft obligation. For use cases incompatible with AGPL (proprietary product or service on top), a separate **commercial license** is available. Contact `hello@foresthub.ai`.

**Q: Is ForestHub free?**
A: Yes, the managed platform is free to start at https://app.foresthub.ai. Or self-host `edge-agents` directly, no account needed.

## Pages

- [Home](https://www.foresthub.ai/en/): Overview of the ForestHub platform
- [Platform](https://www.foresthub.ai/en/platform): Visual workflow builder, engine runtime, deployment paths
- [Edge Agents](https://www.foresthub.ai/en/edge-agents): What edge agents are, edge agents vs cloud agents, and how ForestHub delivers bounded, auditable edge agents
- [Engine](https://www.foresthub.ai/en/platform): The Go-based engine, covering architecture, hardware IO, industrial protocols, LLM routing, deployment, and hardening
- [Patterns](https://www.foresthub.ai/en/patterns): Six graph-first patterns industrial loop-first frameworks can't deliver
- [Roadmap](https://www.foresthub.ai/en/roadmap): What ships today, what comes next, and the longer-horizon vision
- [Solutions](https://www.foresthub.ai/en/solutions): Industrial use cases such as service diagnostics, building automation, predictive maintenance, and autonomous sensing
- [Services](https://www.foresthub.ai/en/services): Edge AI consulting and integration services
- [About](https://www.foresthub.ai/en/about): Company story, team, and mission
- [News](https://www.foresthub.ai/en/news): Product updates and engineering insights
- [FAQ](https://www.foresthub.ai/en/faq): Frequently asked questions
- [Jobs](https://www.foresthub.ai/en/jobs): Career opportunities
- [BoardSmith](https://www.foresthub.ai/en/boardsmith): Open-source KiCad schematic + firmware tool
- [Contact](https://www.foresthub.ai/en/contact): Get in touch

## Reference

- [Glossary](https://www.foresthub.ai/en/glossary): 50 edge AI and agent terms in five categories (foundations, models, agents, safety, industrial), each with a vendor-neutral definition and a permalink anchor
- [Compare](https://www.foresthub.ai/en/compare): How ForestHub compares to adjacent tools, with sourced facts and quarterly verification dates
- [ForestHub vs n8n](https://www.foresthub.ai/en/compare/foresthub-vs-n8n): Graph-first edge agents platform vs general-purpose workflow automation
- [ForestHub vs Node-RED](https://www.foresthub.ai/en/compare/foresthub-vs-node-red): Bounded AI agents for production vs the classic IoT flow tool
- [ForestHub vs Litmus Edge](https://www.foresthub.ai/en/compare/foresthub-vs-litmus): Agent orchestration layer vs industrial DataOps platform
- [ForestHub vs LiteRT-LM](https://www.foresthub.ai/en/compare/foresthub-vs-litert-lm): Orchestration platform vs on-device inference runtime
- [Changelog](https://www.foresthub.ai/en/changelog): Curated stream of user-visible platform updates, RSS at https://www.foresthub.ai/en/changelog/rss.xml
- [Guide: What is Edge AI Orchestration?](https://www.foresthub.ai/resources/guides/what-is-edge-ai-orchestration): Definition and architecture of orchestrating AI workflows on edge devices
- [Guide: Edge Agents vs Cloud Agents](https://www.foresthub.ai/resources/guides/edge-agents-vs-cloud-agents): Where each model fits and why bounded agents belong next to machines
- [Guide: How to Build Agentic Edge AI](https://www.foresthub.ai/resources/guides/how-to-build-agentic-edge-ai): Practical steps from sensor signals to a deployed agent graph
- [Guide: Edge AI Agents on Microcontrollers](https://www.foresthub.ai/resources/guides/edge-ai-agents-on-microcontrollers): What runs on MCU-class hardware and where a Linux gateway takes over

## Contact

- **Website**: https://www.foresthub.ai
- **Platform**: https://app.foresthub.ai
- **Email**: hello@foresthub.ai
- **LinkedIn**: https://www.linkedin.com/company/foresthub-ai/
- **GitHub**: https://github.com/ForestHubAI
- **Location**: Villingen-Schwenningen, Baden-Wuerttemberg, Germany

## Extended Information

For comprehensive documentation, see [llms-full.txt](https://www.foresthub.ai/llms-full.txt).