01
It opens and organizes your project files
PDFs, Word documents, PowerPoint slides, spreadsheets, code, images, and notes appear in one desktop workspace with preview, file tree, and annotations.
Your personal research agent
Hand your research files to an AI that actually does the work.
Atlaslab is a desktop AI agent built for clinicians, scientists, and labs. Drop your PDFs, spreadsheets, papers, and code into one project — the agent reads them, plans the analysis, runs the code, generates the plots, drafts the report, and writes the paper. You direct. It executes.
In plain language
Atlaslab is what ChatGPT would be if it could actually open your files, run your statistics, and finish the boring parts of your research for you.
1Drop in your folder
2Tell it what you want
3Get a finished result
Project files
clinical_data.xlsx
paper_screening.pdf
analysis.py
draft_report.docx
regression.ipynb
figures/
references/
clinical_data.xlsx
analysis.py
regression.ipynb
IDAgeGroupOutcome
00162AImproved
00257BStable
00371ARelapse
00449BImproved
00564AStable
Outcome by group
generated · matplotlib
Agent Chat
GPT-5
statsmodels.logit
running
Ask the agent...
What it actually does
A research agent that opens your files, runs your analysis, and writes your report.
02
You talk to an agent beside the files
Instead of copying text into a chatbot, you ask the agent to read the current file, compare documents, clean a table, run code, or draft a report.
03
It produces reviewable outputs
The agent can create tables, charts, notes, reports, slides, code, and learning cards while keeping sources, tool calls, and intermediate files attached to the session.
04
The website manages accounts, plans, gateway, and market
You download the desktop app here, log in with a website account, use managed model credits through the gateway, and install official Skills and Tools from the market.
Workflow
Designed for research work that must be reviewed later.
Collect sources
Drop your PDFs, spreadsheets, slides, notebooks, code, and notes into one project folder. The agent reads them all.
Plan with the agent
The agent asks clarifying questions, picks the right tools, and follows official Skills tuned for clinical and bioinformatics work.
Run tools
Python, R, SQL, document parsing, figure generation, statistical models — every tool produces an artifact you can open and inspect.
Deliver with trace
Reports, figures, slides, and tables stay linked to their sources and tool calls. Any reviewer can audit any number back to its origin.
Vs other AI agents
Why a research-specific agent beats a general one.
Chatbots answer questions. Coding agents write code. Atlaslab is built end-to-end for research work — it touches your files, runs your analysis, and writes the report, all under one workspace with a full audit trail.
ChatGPT
Codex / Cursor
Atlaslab
Reads your local files
—
Code only
✓
Runs statistical analysis
—
If you script it
✓
Generates plots and figures
Mock only
Via plot libs
✓
Drafts research reports
Plain prose
—
✓
Traces every result to its source
—
—
✓
Built for clinical / scientific work
—
—
✓
Data stays on your machine
—
Some
✓
›
Analyze cohort.csv and draft a methods section.
Reading cohort.csv ...
✓
1,247 rows · 12 columns · 3.2% missing
Running logistic regression ...
Generated forest_plot.png
✓
Drafting methods.docx ...
OUTPUT
forest_plot.png
methods.docx
summary_table.csv
analysis.ipynb
↳ all 4 outputs trace back to cohort.csv
An actual session: one prompt → automatic file reading, analysis, plot generation, and draft writing — every output traced.
Invitation
Invite a collaborator. You both get more model credit.
Generate a personal invitation link from your Dashboard. When someone signs up through it, the reward is added to your gateway quota — and theirs.
Why Atlaslab
Built for teams that need evidence, not only answers.
Local-first workspace
Your files and chat history live on your machine. The agent reads files only when you ask it to — nothing is indexed in the background or shipped to a server.
Traceable execution
Every plan, tool call, intermediate file, and final artifact is recorded in a Data Flow graph. You can audit any number, any figure, any sentence back to its source.
Curated capabilities
Official Skills and Tools are reviewed for clinical, bioinformatics, writing, and learning workflows. One-click install, automatic updates, no DIY required.
Pricing
Start free. Subscribe when you use it daily. Top up when you don't.
The desktop app is free forever. Paid plans give you a generous monthly quota across GPT, DeepSeek, and GLM. Prefer pay-as-you-go? Top up a balance and we'll bill each model call individually.
Free
¥0/mo
Try Atlaslab, no card needed.
- Full desktop app
- 3 trial conversations / day
- DeepSeek Flash only
- Invitation link to earn credit
Plus
¥499/mo
Light personal use.
- GPT · DeepSeek · GLM included
- Monthly quota for casual research
- All official Skills and Tools
- Session reports and Data Flow
Most common
Pro
¥799/mo
Moderate personal use.
- 2× the Plus monthly quota
- GPT · DeepSeek · GLM included
- Priority gateway routing
- Everything in Plus
Max
¥2100/mo
Heavy use for PIs and power users.
- 10× the Plus monthly quota
- GPT · DeepSeek · GLM included
- Highest concurrency and rate limits
- Everything in Pro
Or skip the subscription and top up a balance.
Recharge any amount and we'll bill each model call from your balance — GPT, DeepSeek, GLM, every model priced individually. No expiry, no monthly commitment.
Contact
Deploy Atlaslab for a lab, hospital, or research team.
Talk to us about pilots, official skill publishing, enterprise deployment, gateway plans, and team governance.
Beta 1.0