Most agent frameworks help you build a workflow. The harder part starts after that: the workflow has to run as a long-lived process, fail clearly, restart carefully, and be inspectable after the fact.
That's the gap I'm building AgentOS for — an open-source, Rust-first runtime layer that sits underneath frameworks like LangGraph, AutoGen or CrewAI instead of replacing them.
cargo run -p agentos-cli -- run --agent examples/simple_agent.toml
That single command brings up a supervised agent, a health endpoint, a gRPC message bus, a live SSE event stream, and a recorded trace you can replay later. No API key is needed just to bring the runtime up.
Your agent does something weird on step 7. Reproducing it costs real API calls, and it never behaves the same way twice.
AgentOS journals every LLM exchange and tool result at the provider boundary, so any run can be replayed deterministically — and forked into alternate timelines:
agentOS run --agent my_agent.toml # every step journaled automatically
agentOS replay --session agent_123 # offline re-run, no API cost, drift-checked
agentOS fork --from ckpt_4 --prompt "try the other path"
The dashboard's Recordings view turns those journals into a scrubbable timeline: step through the prompt, each exchange, tool calls and their results, with per-exchange checkpoints as fork anchors.
crates/kernel
— lifecycle, agent handles, supervisorcrates/bus
— in-memory, gRPC, SSE and WebSocket messagingcrates/trace
— recording, replay, diff, checkpoint modelcrates/vault
— secret isolation, encryption, scopes, auditcrates/memory
, crates/registry
, crates/llm
, crates/cli
, crates/sdk
dashboard/
— React debugging surfaceStable enough for local use: the run / ps / logs / trace / replay
CLI flows, local state inspection, export and import, and the core crates with workspace checks and tests.
Still experimental: the dashboard, the WASM plugin runtime, Docker Compose packaging, LLM provider integrations, and Python/TypeScript SDK packaging. Stronger restart and recovery guarantees with explicit tests are next. It is an alpha, and I would rather say that than claim production hardening it has not earned yet.
git clone https://github.com/WAHIB-EL-KHADIRI/AgentOS
cd AgentOS
cargo build --workspace
cargo run -p agentos-cli -- run --agent examples/simple_agent.toml
Repo: https://github.com/WAHIB-EL-KHADIRI/AgentOS
Feedback on the runtime boundaries is what I want most: if you run agents today, what breaks first for you — supervision, observability, or reproducing failures?