By OwkinLearning Objectives
- Understand what an AI research agent for biology does and how it differs from a general chatbot
- Tell the free tier apart from the licensed enterprise tier, and know what each can reach
- Read the vendor's performance claims critically, and know the access and regulatory limits before relying on it
What Is Owkin K Pro?
K Pro is an AI agent for drug research from Owkin, a Paris-based AI company. A researcher asks a question in plain language, such as which tumors express a target gene or which patient subgroup responds to a therapy, and K Pro plans the analysis, runs it against patient datasets and the scientific literature, and returns a written answer with statistics and charts. Owkin calls it an AI scientist for biopharma decision-making.
Under the hood, K Pro is built on OwkinZero, the company's own biological reasoning model, which it describes as a fine-tuned language model trained to reason about biology. The agent orchestrates specialized skills and tools rather than answering from memory, so its answers are tied to the data it was pointed at.
The product has a short naming history. Owkin launched a free research co-pilot called K Navigator in May 2025, then launched K Pro for pharmaceutical and biotech companies in October 2025. Owkin's own site now presents the free tier as K Pro Free, and its old K Navigator page leads there.
⚠️Warning
K Pro is a research tool, not a cleared medical product. Owkin discloses no regulatory clearance for it, and the Owkin connector inside Claude is labeled for research use only. The paid tier is sold for professional and company use, not to individuals, and Owkin says it may restrict access for users whose profession or country applies anti-gift or anti-bribery rules. A pilot with the Gustave Roussy cancer center that puts K Pro in front of molecular tumor boards is a study the partners intend to publish, not an approved clinical use.
📝Note
The free tier and the paid tier see different data. K Pro Free works on 19 public datasets and does not accept uploads. The paid tier adds Owkin's proprietary data, including its MOSAIC spatial and multi-omics oncology dataset, and lets a team bring its own data, which is where most of the product's value sits.
🎯Tip
Explore K Pro: start with the free tier at owkin.com/k-os/k-pro, or book a demo from the same page for the enterprise version.
Pricing
Owkin's own product page lists no prices. It compares a free tier with a Professional tier sold through sales. The dollar figures come from Owkin's listing on AWS Marketplace, where K Pro has been available since November 2025, and are annual contracts.
- 19 public datasets
- No data upload
- Limited usage; activity is logged
- Cloud access only
- One seat
- 12-month contract
- Non-refundable
- Ten seats
- 12-month contract
- Non-refundable
- Owkin proprietary data including MOSAIC
- Upload your own data
- High usage, dedicated account manager
- Cloud or on-premise deployment
The Marketplace listing does not spell out which datasets each seat tier includes, so confirm data access with Owkin before comparing its tiers with the Professional plan on Owkin's site.
Core Features
Plain-Language Analysis
A researcher types a question and the agent turns it into an analysis plan, runs it and writes up the result, without the researcher writing code.
Target and Biomarker Work
Owkin positions K Pro for prioritizing drug targets, assessing whether a target can be drugged safely, validating biomarkers across data types, and characterizing patient subgroups.
Proprietary Patient Data
The paid tier draws on Owkin's multimodal patient data, gathered through its network of hospital and research partners, alongside public sources such as The Cancer Genome Atlas and the scientific literature.
Specialized Agents
In January 2026 Owkin opened its agent infrastructure to outside integration. Its first external agent, Pathology Explorer, reads digitized tissue slides to locate and count cell types, and it launched inside Anthropic's Claude for Healthcare and Life Sciences through a Model Context Protocol (MCP) connector, which requires a K Pro account. Owkin also builds custom agents for licensees; its AstraZeneca agreement includes competitive-intelligence agents.
Strengths
- A free way in — any researcher can try the agent on public data before a company commits to a license
- Answers grounded in data — the agent runs analyses on datasets rather than relying on a chatbot's recall
- Rare proprietary data — the paid tier reaches multimodal oncology data that most research teams cannot obtain themselves
- Real enterprise uptake — AstraZeneca, Sanofi, Boehringer Ingelheim and Servier all signed K Pro licenses between May and September 2026
- Works with other tools — Pathology Explorer is available inside Claude, so a team can use Owkin's agent without switching platforms
Limitations and Considerations
- No regulatory clearance — it is a research and decision-support tool, not a product cleared for patient care
- Professional use only for paid plans — the enterprise tier is sold to companies, and Owkin may restrict access by profession or country
- Expensive at the paid tier — a single seat lists at 100,000 dollars a year on AWS Marketplace, which puts it out of reach for most academic labs
- The free tier is limited — public data only, capped usage, logged activity and no uploads
- Vendor-reported results — claims such as a 20-times productivity gain and 70 percent faster target identification are Owkin's own figures, not independent studies
- Oncology-heavy — the proprietary data is strongest in cancer, so other therapeutic areas get less depth
- Outputs still need scientists — an agent's analysis is a hypothesis to test, not a finding to act on
Best Use Cases
| Use Case | Why K Pro Fits | Caveat |
|---|---|---|
| Early target prioritization | Scores targets against patient data in minutes | Treat rankings as hypotheses to validate |
| Patient subgroup analysis | Finds subgroups across multimodal oncology data | Proprietary data requires the paid tier |
| Literature and gap analysis | The free tier covers public data and papers | Usage is capped and logged |
| Pathology slide analysis inside Claude | Pathology Explorer runs through an MCP connector | Research use only; needs a K Pro account |
| Competitive intelligence for pharma teams | Owkin builds custom agents for licensees | Enterprise contract required |
How K Pro Compares
K Pro answers questions about patients and diseases; it does not design molecules. That separates it from Chai Discovery, which generates antibodies and predicts protein structures, and from Insilico Medicine, which pairs generative chemistry with its own drug pipeline. A team might use K Pro to decide which target and which patients matter, and a design tool to make a molecule against that target.
Key Takeaways
- K Pro is Owkin's AI agent for drug research, built on its OwkinZero biological reasoning model, that answers plain-language questions by analyzing patient data
- The free tier launched in May 2025 as K Navigator and now appears as K Pro Free, limited to 19 public datasets with capped, logged usage
- The paid tier adds Owkin's proprietary oncology data and data upload; AWS Marketplace lists one seat at 100,000 dollars a year
- AstraZeneca, Sanofi, Boehringer Ingelheim and Servier licensed K Pro in 2026, and its Pathology Explorer agent runs inside Claude
- It has no regulatory clearance, and its performance claims are Owkin's own, so use its outputs as research hypotheses
