From your disease to your treatment options, with AI.

OpenEvidence is built for asking medical questions against published evidence and clinical guidelines. For oncologists and other clinicians, that guideline layer is a large part of the value: what does the relevant guidance say, quickly, in the flow of clinic.

Yuga is built around a specific patient’s case. You drop in the chart, reports, papers, and, when they exist, personal multiomics. Agents research this tumor (trials, diagnostics, off-label and frontier paths, atlas context), interrogate the actual figures and tables in what you uploaded, and, when the data supports it, run analysis, not only retrieve a citation or a guideline. The care team shares the same workspace.

Use OpenEvidence when you want to know what the evidence and the guidelines say about a question. Use Yuga when you need a continuously updated, shareable map from this disease to the options in front of you, grounded in the records and the data.

You can. The shape is close: a scientific agent with specialist skills, memory, and the ability to run sophisticated data analysis on high-performance compute.

What it is for is not. Yuga’s skills are deep on patient-specific cancer biology (this tumor, this dataset, this literature, this team) and thin everywhere else. We do not carry a cheminformatics or generative-biology skill set. If the work is designing a molecule or running open-ended discovery, Phylo or Claude Science is the better tool.

Because we support patients, we are also opinionated about compute and abstract the provisioning details away. Interactive analysis runs in a sandbox we provide. Long, repeatable jobs, including raw sequencing, run on HPC we operate. You are not asked to bring your own cluster, or to glue a science agent to a separate collaboration product. Research, analysis, memory, and the care team live in one place.

General chatbots are not built to turn your records into a structured case, read a 100-page report at the level of a figure or a table, run analysis on your sequencing, or sit in a workspace your oncologist can share.

They also are not a health-data product: depending on settings, conversations may be used to improve models, and there is no care-team permission model, no living wiki, and no systematic pass over trials, drugs, targets, genes, or cancer atlases for your disease.

Yuga is case-native: ingest, synthesize, analyze, personalize, continuously updated, cited, and shareable. We do not use your health or genomic data to train models, and we do not sell it.

Privacy PolicyHealth and Genetic Data Privacy Policy

Yuga Bio is a health-data and research workspace for patients and care teams, from a structured case built from your records to a living roadmap from your disease to your treatment options, grounded in literature, trials, diagnostics, and molecular context.

Start with an AI-ready health data vault: upload records, synthesize a structured case, and connect external agents to your files and context. Add the research harness when you want in-app agent threads for trials, literature, analysis, and a shared wiki with your care team.

Patients and caregivers who want health records turned into a structured case and, when ready, a rigorous pass over trials, literature, and options beyond standard of care.

Families who want an AI-ready vault first and can add full research later, without rebuilding from scratch.

Clinicians who want an integrated, shareable picture of a patient's case they can interact with.

Scientists who need patient-specific cancer biology on a particular tumor, not a general discovery agent.

You can start before any molecular workup, after a first panel, or once you have deep tumor profiling. The workspace grows with whatever you have.

No. A chart, pathology report, case note, or even a diagnosis is enough to build a structured medical record, research trials, and literature.

When molecular results arrive, the same workspace becomes grounded in your variant calls or biomarkers. After deep tumor profiling, you can bring personal genomics datasets including raw sequencing and have it analyzed here, next to integrated molecular atlas datasets. You do not wait for a perfect dataset to leave fewer stones unturned.

Yuga Bio's agent, Krita, comes with cloud storage and compute that is private to you. It can understand where you files live, ingest them from your Dropbox, Google Drive or AWS object storage. Interactive analysis on the datasets you drop in runs in a sandbox, including custom bionformatics on sequencing data, as well as traditional data science work on automatically structured tables and figures from your documents.

Long, repeatable jobs on large files, including processing raw sequencing, run on high-peformance compute we provision and operate. Analysis results are synthesized into cited reports with figures and tables, easily shared via download or email.

No. Yuga is a research and hypothesis-generation workspace. It is not a medical device, is not FDA-cleared, and does not diagnose, recommend, or decide treatment.

Every output is meant to be opened, checked against the cited sources, and reviewed by a qualified clinician who knows the full clinical picture. Use of Yuga does not create a physician-patient relationship with us.

Terms of Use

Yes. Collaboration lives in the same workspace as the research. A patient or authorized caregiver invites clinicians, scientists, and other caregivers into the threads, datasets, wiki, and case memory the agent is already using. Teammates can pick up where you left off by uploading data, or advancing a research thread without a separate handoff.

You can also export or email reports and briefs as PDFs. You control who is on the team and what they can see.

Electronic health records, PDFs, DOCX, and PPTX; pathology and imaging reports; lab panels; case notes; research papers; personal genomics files and raw sequencing data when you have them.

Records and papers are not flattened into a summary. You can ask about a figure, a table, a specific dated note, a list of variants from a vendor PDF, and get an answer grounded in that artifact and in the case. Sync from Google Drive or Dropbox is available on supported plans.

Vault ($20/mo) is an AI-ready health data vault: upload records, synthesize a structured case, and connect external agents without an Anthropic key. BYOK Harness ($50/mo) and Pro Harness ($100/mo) add the full in-app research agent, wiki, integrations, and parallel threads. Practices and institutions serving multiple patients can inquire about a custom plan with tailored data use rights, including a HIPAA business associate agreement where that channel applies.

See current plans on the homepage, or email us if you need something tailored.

PricingInquire about a custom plan

The service is offered to individuals who are residents of, and physically located in, the United States.

No. We do not train foundation models, and we do not fine-tune a Yuga model on your cases. We focus on building the harness: high-quality data sources, connectors into records and labs, compute environments, tools, and skills that makes an agent reliable on a personalized tumor-derived treatment ladder.

That is a different question from whether we use your data to improve models. We do not use health or genomic data to train, fine-tune, or improve models. See Privacy.

Do you use my data to train models?

Most open harnesses are general loops: tools, code, memory, a filesystem. That is a strong way to write software. It is a weak default for N-of-1 cancer biology on health data.

Yuga’s harness is opinionated. Data sources, connectors, sandbox and HPC, tools, and skills are chosen for a patient’s workflows. A living case file and care-team collaboration sit in the same loop. You can point a general harness at a folder of PDFs; you will still have to build trial-grade sources, figure-level paper reading, sequencing compute, and a workspace an oncologist can actually join.

We continue to leverage the best ideas from teh fast moving field of harness engineering (context management, token efficiency, model routing, and memory) to ensure a longitudinal case, with hypotheses for treatments and tests, a library of releveant papers and trials, bioinformatics analysis results of personal genomics, and an entire team's context, stays coherent and useful over time.

We run frontier model, primarily Anthropic's Claude Opus tier. We do not train a Yuga foundation model.

Vault includes Yuga-paid case synthesis without an Anthropic key. On BYOK Harness you bring your own Anthropic key and pay the provider directly. Pro Harness includes Yuga-billed model usage metered against monthly credits.

Pricing

Skills are packaged ways of working the agent can retrieve and use on demand: how to interrogate a variant, walk a trial record, pull a figure from a paper, kick off a sequencing job. We ship a large library aimed at patient-specific cancer biology.

You can bring your own skills on supported plans, alongside the ones we maintain.

Search is the start, not the product. Documents and papers are decomposed into the artifacts that actually carry the science (figures, tables, equations, methods) so the agent can answer a detailed question against that panel or that row, then follow up on its own, tied back to the case.

A single embedding lookup over full text is not enough for a pathology deck or a supplementary figure. That is why reading sits in the harness, next to the sources and the tools, rather than as a vision-language model call around a PDF.

Yes. You can delete your account directly from within the app. We purge your uploaded documents and data, your chat history, and any AI-generated datasets or insights, at the scope of your entire team.

We may retain billing and usage history to resolve disputes or enforce our agreements.

Privacy PolicyHealth and Genetic Data Privacy Policy

We do not use your data to train, fine-tune, or improve AI models except with your separate, explicit, opt-in consent.

We do not use genomic sequence data or other identifiable health information to train, fine-tune, or improve AI models under any circumstances.

Privacy PolicyTerms of Use

No. We do not sell or rent your data to anyone, including pharmaceutical companies. We do not use consumer health data for targeted advertising, and we do not share it with advertising networks, data brokers, or analytics providers for cross-context behavioral advertising.

Secondary research use of consumer health data happens only with a separate, express, opt-in consent. We do not disclose genetic information to employers or to health, life, long-term-care, or disability insurers without your consent.

Health and Genetic Data Privacy Policy

You, and the people you invite onto your care team. Shared threads, wiki, datasets, and outputs are visible to team members at the team owner’s discretion.

Processors (cloud infrastructure, support, security), and contractors see data only as needed to run the service, under written contract. We may disclose data if required by law.

Privacy Policy

If you use Yuga directly as a patient or caregiver, we are not acting as a HIPAA covered entity or business associate for that channel. State consumer-health and genetic-privacy rules still apply; those are in the Health and Genetic Data Privacy Policy.

If a health-care provider or other HIPAA covered entity uses Yuga on behalf of patients, that channel is under a business associate agreement. Custom plans for practices and institutions can include a BAA.

Health and Genetic Data Privacy PolicyCustom plans

We de-identify all uploaded documents at the software level before the AI agent works with them. We currently support PDF, DOCX, PPTX, JPEG, and PNG.

Raw PHI documents are stored in a sequestered location inaccessible to the AI agent. By design, we do not send PHI to LLMs. De-identification is not perfect, and we treat residual risk accordingly.

Privacy PolicyHealth and Genetic Data Privacy Policy

Agent Krita can search the literature down to figures and tables, clinical trials, diagnostic products, patient forums, drug labels, targets, genes, pathways, and human cancer omics atlases, plus the private health records, lab reports, and genomics files you drop in. Exact coverage will keep expanding.

Yuga often shines relative to single-pass search tools by combining code-writing capabilities to implement custom prioritization frameworks applied to results from structured queries from these sources to retrieve and prioritize information in multiple steps.

Yes. Full-text and abstract search across open biomedical literature, including PubMed, arXiv, bioRxiv, and medRxiv from multiple agent-ready data providers.

Yes. Structured trial search from 13 global registries.

Agents can look up labeled and investigational therapies in the context of the case, including where evidence is off-label or still emerging.

Outputs are hypotheses with sources, not prescriptions. Nothing here is a recommendation to start, stop, or switch a drug.

Yes. When the case includes variants, expression, or a gene of interest, agents can ground options in the gene and the target: which cellular compartments is it expressed, what the drug development landscape looks like against a given target, what pathways are impacted in the tumor, how they could be exploited for off-label combination strategies, and more.

That sits next to your molecular results when you have them, not a generic gene page in another tab.

Yes. Cohort-scale molecular resources such as TCGA, used to put a tumor in the context of a matched cohort: similar profiles, pathway patterns, and what that has meant in published cohorts.

Atlases provide much needed context for patient's data and often help reason about options for undiagnosed, rare, or progressing cancers.