High-performance infrastructure for AI agents.
We build the fast parts of the AI stack. Rust-accelerated drop-in replacements for the tools agents actually depend on — LLM gateways, workflow engines, package managers, and developer tooling. Zero config. No rewrites. Just faster.
29 public projects · MIT/Apache 2.0
🌟 Regulus · site ↗ — The EU & UK compliance plane for Google ADK.
Runtime BasePlugin suite encoding 10 regulations (EU AI Act, GDPR, DORA, NIS2, EHDS, UK GDPR, FCA SYSC, PRA SS1/23, PRA SS2/21, NHS DSPT) and 6 governance frameworks as runtime profiles. Hash-chained audit envelopes, PII redaction, dual-control kill switch, fail-closed data residency.
Java 21 · Maven · regulus-ai-adk-plugins · Repo · site ↗
The agent stack is built on Python libraries that were never designed for production concurrency. We profile, rewrite the hot path in Rust, ship as a drop-in replacement. Your code doesn't change — it just gets faster.
| Project | What it accelerates | Speedup (per the project) |
|---|---|---|
| fast-litellm · site ↗ | LiteLLM connection pooling, rate limiting | 3.2× pool, 2.8× rate limit |
| fast-langgraph · site ↗ | LangGraph checkpoint + state management | 700× checkpoint, 10–50× state |
| fast-crewai · site ↗ | CrewAI serialization, tools, DB search | 34× / 17× / 11× |
| fast-axolotl · site ↗ | Axolotl LLM fine-tuning (dedup SHA256) | Multi-threaded, no OOM |
| fastworker · site ↗ | Python background workers | No Redis. No RabbitMQ. |
| ormai · site ↗ | Agent DB access (LangGraph / PydanticAI / SQLAlchemy / Tortoise) | Safe text-to-SQL |
| fastagentic · site ↗ | Production runtime for PydanticAI / LangGraph / CrewAI / LangChain | Cached run_id, MCP server |
| Project | Description |
|---|---|
| mcp-pay · site ↗ | Payment awareness layer for MCP (Model Context Protocol). |
| agentvfs · site ↗ | Workspace runtime and execution boundary for AI agents. |
| brat · site ↗ | Multi-agent harness for AI coding tools. Crash-safe state, parallel execution, one CLI. |
| ringlet · site ↗ | One CLI to rule all your coding agents. Profile-isolated, cost-tracked. |
| grite · site ↗ | The issue tracker that lives in your repo. Built for AI agents. |
| memorg · site ↗ | Give your LLM a memory that actually works. |
| stratafs · site ↗ | A semantic filesystem that turns file storage into a searchable knowledge base. |
| openclawOS · site ↗ | OS-like architecture for AI assistants with kernel-based, process-isolated applications. |
| openclawMU · site ↗ | Multi-tenant fork of OpenClaw with strict data isolation. |
| openclaw-rs · site ↗ | Community Rust implementation of OpenClaw. |
| ukkin · site ↗ | Create AI agents on your phone that automate your daily tasks. |
| m9m · site ↗ | The n8n alternative without the bugs — faster, more reliable workflow automation in Go. |
| Project | Description |
|---|---|
| gity · site ↗ | Make large Git repositories feel instant. |
| stout · site ↗ | Stout is a drop-in replacement for the Homebrew CLI that's 10–100× faster for most operations. |
| stout-index | Pre-computed package index for stout · site ↗. |
| rjest · site ↗ | A blazing-fast Jest-compatible test runner. 100× faster warm runs. Zero config changes. |
| rpytest · site ↗ | Run your pytest suite faster. Change nothing. |
| rninja · site ↗ | Build faster. Cache smarter. Drop-in ready (Ninja replacement). |
| recurl · site ↗ | curl that just works. Drop-in replacement with automatic anti-bot bypass. |
| rewget · site ↗ | wget, but it works everywhere. |
| stkd · site ↗ | Stacked Diffs for GitHub and GitLab (Graphite-compatible). |
The AI agent stack is built on Python libraries that were never designed for production concurrency. LiteLLM's connection pool uses stdlib locks. LangGraph checkpoints serialize through a single thread. Package managers shell out to Ruby. The result: agents that work great in a demo and fall over in production.
We fix these bottlenecks one at a time: profile, rewrite the hot path in Rust, ship as a drop-in replacement. Your code doesn't change. It just gets faster.
Every project is MIT or Apache 2.0 licensed and accepts contributions. See the individual repository for setup instructions.
Neul Labs builds high-performance, open-source infrastructure for AI agents — Rust-accelerated drop-in replacements for the tools agent stacks depend on. Founded to fix the production bottlenecks in the Python AI stack. All projects MIT/Apache-2.0.
- 🌐 Website: https://www.neullabs.com
- 📚 Docs: https://docs.neullabs.com
- 💻 GitHub: https://github.com/neul-labs
- ✉️ Contact: contact@neullabs.com