Work with your data
Five new built-in skills turn a data file into results you can report. You don't need to call them by name; ask in your own words and the right one loads.
- analyze answers questions from a CSV, TSV or JSON file: "what does my data show?"
- explore-data profiles and cleans a dataset, including SPSS, Stata and R files, and saves a cleaned copy beside the original.
- statistical-analysis picks the right test, checks its assumptions, runs it, and reports the effect size with its confidence interval.
- create-viz makes the right chart for your question, ready for a paper.
- validate-data checks a draft against its data before you submit: every number, table and chart recomputed, and every mismatch listed.
Systematic reviews
The new systematic-review skill helps with a systematic or scoping review: a protocol, repeatable searches, screening, a PRISMA diagram, data extraction, risk of bias and meta-analysis.
Make your own skills
Ask Multilo to "turn what we just did into a skill" and the skill-creator skill writes it, tests it and improves it. Type /init and Multilo looks around your project and writes an AGENTS.md describing it, so every chat starts knowing how the project works.
Bring your memory from another assistant
In Settings → Memory, click Import from another assistant. Multilo gives you a prompt to run in ChatGPT, Claude or Gemini, reads what you paste back, shows you what it plans to remember, and saves it only when you say yes. Sensitive details like health or money are left out, and instructions hidden in the paste are never followed.
Auto uses fewer credits
When the model is set to Auto, Multilo now uses the efficient model unless your task needs a stronger one, and keeps one model for the whole chat. Long chats also reuse more of what was already sent, so the same work costs less.
See what fills the chat
/context, or the ring beside the chat box, now shows a coloured bar of what the chat is holding (instructions, tools, skills, subagents, messages) and how much room is left before it compacts, with tips when something is taking too much. It works in every chat, and after a restart.
Also in this update
- Full author names: literature search results show authors' full names.
- Multilo now needs macOS 13 (Ventura) or later, as it moves to a newer, supported version of Electron.