StateFabric a small infrastructure layer for AI agents that need more context than chat history.

Once an agent starts using tools, running for longer sessions between restarts, ‘keep the messages in memory’ doesn’t really cut it. You need to know:

- what happened?

- what state changed?

- which tools were called?

- what context should go into the next model turn?

StateFabric stores an append-only event log for agent runs, then derives working context from it.

Today it supports:

- Durable sessions and event storage

- User/model/tool event timelines

- Reconstructed session state from stored events

- Compacted model-facing context

- A dashboard for inspecting sessions, raw payloads, compaction artefacts, and usage

- Google ADK integration via @statefabric/adk

- Direct Node/REST usage via @statefabric/client (for custom runtimes)

It’s not trying to be an agent framework. Your runtime still owns prompts, tools, model calls, and business logic. StateFabric sits beside it as the state, replay, and context layer.

It's an early launch, so I’m deliberately starting with narrow integrations. ADK (JS) and custom node runtimes initially.

I’d love feedback from people building agents that need persistence, replay-ability, or better context management across long-running sessions.