Our products

ldgr

Live

Durable work history for AI agents.

ldgr records what an agent did, what changed, and why. Tasks, runs, evidence, and decisions live in one local SQLite database. The record survives restarts and context resets.

ldgr is built in Rust and ships as a single binary. It wraps agent work in a simple loop: create a task, run it, record what happened, and decide what comes next. It adds a dependency-aware work queue, a prompt history, and a local web cockpit for oversight. Adapters extend it into code, research, and security workflows.

Cue-based memory for AI agents.

Ecphory is a memory system that lets an agent learn and remember as it works.

Ecphory uses context cues to retrieve memories from situations the agent has met before. It stores new situations as it works. It consolidates repeated encounters into one episode, so each new encounter carries what the earlier ones taught. That memory informs what the agent does next. This is not RAG. There is no public repo and no public beta yet. Today ecphory runs as an external module that mounts onto existing LLM services. We are working to move it directly into the model architecture.

AI workers that stay on your hardware.

Aegyx runs AI agents on your own servers, with scoped permissions so each agent can only reach what its task needs. Every action is recorded, so you can always show exactly what the system did.

Aegyx is built for one problem: putting AI workers on sensitive work without giving up control. Your data never leaves your network, and agents get the permissions their task requires and nothing more, so a mistake cannot reach anything it should not. The full action history is auditable, which shortens security review and protects you in front of auditors and customers. The result is less time defending AI to your risk team, and no breach cleanup to pay for. Aegyx is a complete prototype, and it is not generally available yet.