We build AI systems — and the products around them.
DeepFounder AI is an independent lab. We research how AI agents work in practice — memory, tooling, coordination — and ship the software that comes out of it.
Applied AI research
We study how agents behave outside the demo — long-running memory, tool use, coordination between systems. Findings ship as working software, not slideware.
Own products
Agent infrastructure we run ourselves: memory services, issue tracking for agents, self-hosted assistants. Built in Rust and Python, released in the open.
Software development
Custom AI and business-intelligence systems for teams that need them to run on their own hardware — a laptop, a workstation, or their own server.
castor
A self-hosted AI agent built to drop into business workflows — customer ops, internal automation, knowledge retrieval, scheduled reporting. Deploys on a laptop, a workstation, or your own server.
mnemos
Cloud memory for AI agents — a persistent wiki with sources and an index, exposed over REST, MCP, and a CLI.
LineAgent
An issue tracker built for AI agents — REST, MCP, and CLI over a single SQLite file. Agents file, triage, and close their own work.
llm-wiki
An LLM-maintained personal wiki for Claude Code. An active librarian: it ingests, synthesizes, cross-links, and reminds.
Research becomes software
Every line of inquiry ends in a running system. If a finding can't survive deployment, it isn't finished.
Self-hosted by default
Our systems run on your hardware — no mandatory cloud, no data leaving the building unless you send it.
Open by default
Code, interfaces, and methods are public. We keep the weights of our attention on problems, not on secrecy.
Small and exact
A small team, plain interfaces — REST, MCP, CLI — and boring, dependable storage. Precision over surface area.