Serious work, in any discipline, with an AI Assistant that shows its work.
Insight LM is a desktop workbench where you organize source material into workbooks, run analyses in notebooks, and work alongside an AI Assistant that does the work — searching, reading, computing, and building artifacts with deterministic tools over your data. Engineering, legal analysis, education, research — the workbench is the same; your discipline shapes it.
Read the documentationLearn moreWhat is Insight LM?
Insight LM is a lightweight desktop application for workbook management and AI-assisted analysis in any discipline — it is used today for systems engineering, legal work, grading model-based coursework, and financial research. It brings your documents, data, notebooks, and a working AI Assistant into one workbench — inspired by the layout and extensibility of a modern code editor, but built for knowledge work.
Data, not just documents
Documents are views of underlying data. Workbooks organize sources, references, and generated artifacts so results stay traceable.
Notebooks are the engine
Analyses run in real notebooks with managed Python environments, so every result is reproducible and inspectable.
A real Assistant, not a chatbot
The Assistant does the work — it searches, reads, queries, computes, and builds artifacts through deterministic tools over your workbooks, with receipts on its claims.
Capabilities
Workbook management
Import, organize, and version source material. Search and read across a workbook yourself or hand it to the Assistant.
An Assistant over your data
Ask for answers or hand off whole tasks, grounded in your sources. Results cite where they came from.
Notebook execution
Run Python notebooks against workbook data in app-managed environments — no local setup required.
Extensible workbench
Extensions contribute views, commands, tools, and workflows through a stable public API.
Deterministic tool servers
Capabilities are delivered by standalone tool servers (MCP), keeping the assistant's actions auditable and repeatable.
Graphs at the core
An in-app ontology and SPARQL service, graph files as first-class workbook documents, and write-once versioned graph publishing. Objects are created through structured methods, so consistency, provenance, and traceability are properties of the system.
Imperfection, managed
The Assistant is deeply knowledgeable but not infallible — so its work is governed by executable mathematics, scripts, and workflows that check results deterministically instead of taking them on faith.
Architecture
Insight LM follows an extension-workbench architecture: a host application owns the shell, layout, and state; features arrive as extensions through a public API; and AI capabilities are grounded by deterministic tool servers working over structured knowledge graphs — that grounding is where provenance and traceability come from.
Where this is going
In active build — published because it drives the architecture that already ships.
Full work capture in the graph
The whole decision loop — event, assumptions, data, models, simulation and analysis, decision, and back to event — captured in the graph as it happens. “Why did we decide this?” becomes a query, not an archaeology project.
The graph as system of record
Documents, views, and dashboards become projections of graph-held knowledge — no more dead documents drifting from the truth. The workspace grows as your graphs grow.
Graph-first extensions
New extensions build on graphs as their base — starting with lightweight systems modeling where the model is a graph in your workbook and the Assistant is a first-class modeler.
Provenance-aware automation
We run our own CI with these same graph features — builds, verifications, and automations recording provenance into the graph, so the system integrates and verifies its own capabilities.
A system that extends itself
Extensions contribute agents, skills, and workflows through the public API — and the Assistant can author new ones using the same structured, human-gated methods. With hot-loading in build, the workbench moves toward governed self-evolution.
Screenshots & demos
Documentation
Getting-started guides, concepts, architecture notes, and reference material live on the documentation site.
Contact & about
Insight LM is under active development. For inquiries, contact the team at contact@insightlm.dev.