Urika: Letting AI Agents Run Your Exploratory Analysis
Urika is an open-source multi-agent platform that automates exploratory data analysis. You bring a dataset and a question — a team of 12 AI agents does the rest.
My team and I spend a lot of time on exploratory data analysis. Not the final, polished analysis you put in a paper — the messy stage before that. Trying different approaches, comparing methods, figuring out what questions the data can actually answer. It’s essential work, but it’s repetitive and time-consuming. You run dozens of variations on similar pipelines, keep track of what you tried, and manually document results that mostly don’t go anywhere.
I built Urika to automate that stage.
What It Does
Urika is a multi-agent scientific analysis platform. You give it a dataset and a research question, and it orchestrates a team of 12 specialised AI agents that systematically explore your data. A planning agent designs the analytical approach. A task agent writes and executes Python code to implement it — every method is built fresh for your specific dataset and question, not pulled from a template. An evaluator scores each result against your success criteria. An advisor decides what to try next. A literature agent searches relevant papers you’ve added to the project. When the planning agent identifies a capability gap, a tool builder creates custom analysis and modelling tools on the fly — anything from domain-specific data loaders to specialised statistical measures. And when it’s done, a report agent writes up the findings, a presentation agent builds slides, a project summarizer keeps a unified narrative of progress and findings, and a finalizer packages everything with standalone reproducibility scripts.
Think of it as delegating the grunt work of exploratory analysis to a research team that never gets tired and documents everything automatically.
Why I Built It
The honest answer: I was tired of writing the same analysis boilerplate across projects. Profile the data, check distributions, try a few models, compare results, write it up. Every project follows roughly the same loop, but each one requires fresh code and fresh documentation.
I wanted a system where I could describe what I was investigating and have the exploration happen systematically — trying approaches I might not have considered, keeping a rigorous log, and giving me a ranked leaderboard of everything that was tried so I could focus on the results that actually mattered. The multi-agent architecture was a natural choice. After over 6+ months of building and working with agentic systems, I could see that the different stages of a research workflow — planning, execution, evaluation, advising — mapped cleanly onto specialised agents. Urika is really just all of my current AI-assisted workflows rolled into one complete analysis system.
Handling Sensitive Data Responsibly
If you work in academia or with industry partners, you deal with ethical constraints around data constantly. Student submissions, unpublished research, ethics-approved participant data, industry datasets under NDA — there’s always something that shouldn’t be sent to a cloud API. This isn’t just about privacy preferences. It’s about ethics board requirements, consent agreements that never anticipated cloud AI, and data-sharing agreements that restrict where processing can happen.
Urika takes this seriously. It has three privacy modes — but the one that matters most is hybrid mode. In hybrid, your data is only ever read by private endpoints — models running locally on your machine via Ollama, on a secure institutional server, or on your organisation’s own cloud infrastructure. Whatever sits within your data governance boundary. The agents doing the conceptual work — planning experiments, evaluating results, advising on next steps — can still use frontier cloud models where you need the extra capability. You get the power of the best available models for reasoning and strategy, without your sensitive data ever leaving your approved infrastructure.
You can also run fully private (everything on local or institutional endpoints) or fully open (everything cloud) depending on the project. The routing is per-agent, so you have fine-grained control over exactly which models handle which parts of the workflow — and different projects can have completely different privacy settings.
Two Ways to Work
That systematic exploration can run at whatever pace suits you. In guided mode, Urika runs one experiment at a time. You review the result, read the evaluator’s assessment, and decide what to try next. It’s collaborative — the AI does the heavy lifting, but you steer.
In autonomous mode, the advisor agent takes over. It reviews results, identifies gaps, and launches follow-up experiments without waiting for you. You come back to a complete project book with a leaderboard of every method attempted. Urika also connects to email, Slack, and Telegram, so you can get progress updates and even pause, stop, or query a run remotely — useful when you’ve kicked off an autonomous session and stepped away.
You can drive all of this through whichever interface fits the moment. Urika ships with three first-class interfaces that share the same project state on disk: a scriptable CLI for batch jobs, scripts, and CI (every command supports --json output for custom pipelines); an interactive TUI with tab completion, slash commands, and orchestrator chat for exploratory work; and a browser dashboard for monitoring long runs, sharing results, and managing settings. Anything you do in one shows up in the others.
Where It Is Now
Urika is open source (Apache 2.0), early-stage, and in active development. It runs on Linux, macOS, and Windows 11. The current release uses the Claude Agent SDK and requires an Anthropic API key plus the Claude Code CLI — per Anthropic’s Consumer Terms, the Claude Agent SDK that Urika depends on can only be authenticated with an API key, not a Pro/Max subscription. Set a monthly spend limit in the Anthropic console before you start.
That will broaden soon. Adapters for the OpenAI Agents SDK, Google ADK, and PI are on the roadmap. The goal is for Urika to work with whatever models and infrastructure you already have.
If you work with data and want to speed up the exploratory phase, give it a try. I’d genuinely appreciate feedback — bug reports, feature requests, or just hearing how it works on your data.
GitHub: github.com/xkiwilabs/Urika
Urika is part of the xKiwiLabs open-source toolkit. Everything is free, built in the open, and designed for researchers and practitioners who want AI tools that actually fit their workflow.
— Michael Richardson Professor, School of Psychological Sciences Faculty of Medicine, Health and Human Sciences Macquarie University
AI Disclosure: This article was written with the assistance of AI tools, including Claude. The ideas, opinions, experiences, and workflow described are entirely my own — the AI helped with drafting, editing, and structuring the text. I use AI tools extensively and openly in my research, teaching, and writing, and I encourage others to do the same.