BeamWeaver: AI agents and durable LLM workflows for Elixir

We’ve been building LLM features in Elixir and kept running into the same problem: the available Elixir libraries were useful, but still fairly simple, and there was almost no observability for agentic systems.

So we either had to keep Python microservices around for LLM workflows, or accept much less tooling on the Elixir side. We wanted to stop doing that.

BeamWeaver is our attempt to bring the best ideas from LangChain, LangGraph, and Deep Agents into Elixir, with an OTP-native design:

• agents and tool calling
• graph workflows
• checkpoints and resumable execution
• memory stores
• retries, fallbacks, interrupts, and human review
• typed streaming events
• provider adapters for OpenAI, Anthropic, Google Gemini, xAI, and Moonshot/Kimi
• fake/replay models for deterministic tests

We’re also building observability on top of it through WeaveScope, which we’ll release very soon.

The goal is to give Elixir teams the full harness needed to build advanced agentic systems without pushing the hard parts into Python services.

Documentation: weavescope.gitbook.io/beam_weaver

Hex: beam_weaver

GitHub: GitHub