We built our own workflow engine. IronFlow is a Rust-native, DAG-based orchestration engine with 96 built-in nodes that ships as a single binary — no Python, no Docker, no dependency hell. Pair it with IronCrew for multi-agent reasoning and TinyCrew for web-integrated agent workflows. When your team can't use n8n or Dify, we have the alternative.
Three in-house Rust and TypeScript tools for workflow orchestration, multi-agent reasoning, and enterprise integration.
In-house — Rust Workflow Engine
DAG-based workflow orchestration with 96 built-in nodes — HTTP, files, S3, databases (SQLite, ArangoDB), AI embeddings/chunking/chat, MCP client, document extraction, image processing, email/Slack notifications, and more. Define workflows in Lua scripts, execute on a sandboxed async Rust runtime. Parallel step execution, retry with exponential backoff, conditional routing, subworkflow composition, and a built-in REST API. Single binary, zero dependencies, runs anywhere including air-gapped environments.
In-house — Rust Multi-Agent Runtime
When workflows need reasoning, not just sequencing. IronCrew orchestrates multiple AI agents with dependency graphs, agent-to-agent messaging, shared memory with TTL and relevance scoring, collaborative discussions, and structured JSON output. Define agents and tasks in Lua, execute as a single binary. Provider-agnostic — OpenAI, Groq, Ollama, Azure.
In-house — TypeScript Multi-Agent Framework
Bun-native agent orchestration for web applications and APIs. Shared memory, structured output via Zod schemas, multi-model routing for cost optimization, agent-to-agent messaging, conversation history with auto-summarization, and response streaming. When your automation lives inside a web service, not a CLI.
Concrete workflows we build and deploy — not theoretical capabilities.
Anthropic's Model Context Protocol is our preferred standard for connecting AI to external systems. IronFlow includes a native MCP client node.
When the project calls for existing platforms instead of our toolchain.
Production deployments and real IronFlow pipeline patterns.
WOLF GmbH — Multi-Tool Enterprise Assistant
Intent analysis with dynamic routing, multi-LLM fallback chains, tool-calling for data processing, and complex document workflows. 2.5+ years in production handling real enterprise workloads daily.
In-house — Visual Digital Forensics Automation
Multi-LLM video and image analysis pipelines for law enforcement. Automated content classification, threat detection, and evidence flagging with cryptographically verifiable processing chains. Edge-based execution on standard mobile hardware — no cloud upload of sensitive footage required.
IronFlow — Document → Vectors → Search
End-to-end pipelines in a single Lua script: extract content from VTT/PDF/DOCX, chunk with configurable strategies, embed via OpenAI or Ollama, store in S3 Vectors or ArangoDB, then query with semantic similarity and metadata filtering. Full chunk → embed → store → query cycle with parallel step execution.
IronFlow — VTT → Sentiment → Insights
Extract interview transcripts, run sentiment analysis via multi-provider LLM calls (OpenAI, Azure, OAuth endpoints), normalize results across response formats, and generate per-speaker sentiment reports. Handles OAuth token exchange, provider-agnostic chat completions, and structured JSON output in a single automated flow.
Contact us to discuss how IronFlow, IronCrew, and our multi-agent frameworks can automate your workflows — from simple data pipelines to complex agentic systems.
Get in Touch