Your AI provider · Your data · Open source

Research without
losing context.

The Integrated Research Environment (IRE) is a desktop application built for researchers. Your code, literature, experiments, and AI live in one unified workspace — so you never start from scratch.

IRE interface — five-pane research environment with Focus, Resources, Chat, Notes, and Ideas panels

ML research workflows are fragmented.

Context loss

Every session starts with re-establishing state. What did I try? What failed? Why?

Knowledge fragmentation

No persistent memory of insights, paper summaries, or rejected approaches — so AI suggestions repeat dead ends.

Experiment limbo

Fire off a long run, switch tasks, and forget about it. Results sit unreviewed.

Goal drift

The core research question gets buried under technical details and literature exploration.

Everything in one unified workspace.

Persistent LLM Wiki

A Git-tracked markdown wiki maintained automatically. Architectural decisions, failed approaches, and current blockers are selectively injected into the agent's context so it never forgets.

Non-blocking Experiments

Fire off a long-running experiment and keep working. IRE monitors it as a detached process. When it finishes, the agent wakes up with full results and context attached, ready to continue.

Agentic IDE, Deeply Integrated

IRE wraps coding agents with a native MCP bridge. AI agents can read your wiki, write findings, record memory, update your research pulse, and start experiments — not just chat.

Structured Memory

Long-term memory for durable insights alongside short-term daily notes automatically injected into context.

Your Research Stays Yours

IRE never collects or stores your research content. What leaves your device is only what you send to your AI provider — nothing else. Git is your backup and collaboration layer.

One Workspace, One Repo

Each IRE workspace maps 1:1 to a Git repository. Your code, wiki, experiment logs, and AI state all live together in .ire/. Version-controlled from day one.

The five-pane research interface.

Every panel is purpose-built. Nothing is noise.

Focus

Your research question and weekly goal, always visible.

Resources

Indexed papers, articles, and API docs. Submit a URL or file and the agent summarises it into the wiki.

Chat

Streaming coding agents with real-time tool calls, experiment status, and full wiki access. Supports multiple tabs.

Notes

Editable markdown notes that the agent can read and write. Click to inline-edit.

Ideas

A draggable bullet list for capturing hypotheses and next steps.

Free. Open Source. Runs everywhere.

Built with Tauri and a Rust backend — lightweight, native, fast.

Or build from source — Rust + Node.js required.