AIM Research
Trialibre
Beta · Open Source · Free · MIT License

Clinical trial matching for everyone.

Trialibre helps clinicians and researchers find clinical trials for their patients. Upload a patient description, and Trialibre checks it against trial eligibility criteria — one by one.

Trialibre interface showing patient matching
12
Therapeutic areas covered
5
Languages supported
4
LLM providers
8
Input formats

Built for real clinical workflows

Every feature designed around the needs of coordinators, investigators, and research teams — not just engineers.

🔍

Patient Matching

Type, upload, or photograph a patient record. Trialibre checks each eligibility criterion individually and shows which are met, not met, or need verification.

📄

Protocol Upload

Upload your own protocol (PDF, DOCX, or paste text) and match patients against it. Inclusion and exclusion criteria are extracted automatically.

🌐

ClinicalTrials.gov Import

Enter an NCT number to import a trial directly from ClinicalTrials.gov with full eligibility criteria and site locations.

🔒

Privacy First

Built-in de-identification strips patient names and IDs before cloud processing. Or run fully offline with a local AI model — nothing leaves your device.

🌍

Multilingual

Interface available in English, French, Portuguese, Spanish, and Arabic. Patient notes in any language are automatically detected and processed.

🤖

LLM Agnostic

Works with Claude, GPT-4, Llama via Ollama, or any OpenAI-compatible endpoint. Switch providers without changing anything else.

📋

Batch Screening

Submit up to 100 patients at once via the API or UI. Each patient is matched against all loaded trials with per-patient strong/possible/unlikely counts.

📨

Referral Tracking

Generate referral records with patient, trial, and recipient details. Track status lifecycle (created → sent → accepted/declined). Delivery integrations are on the roadmap.

📊

Evaluation Framework

Standard IR metrics (Precision@K, NDCG, MRR, strength accuracy) with 24 annotated patient-trial pairs across 12 therapeutic areas for regression testing match quality.

Three steps to a match

1

Describe Your Patient

Type or paste a clinical note, or upload a PDF, DOCX, CSV, FHIR bundle, or HL7 message. Photo-to-text via Tesseract is available when the system dependency is installed.

2

AI Checks Each Criterion

Trialibre evaluates every inclusion and exclusion criterion individually, with plain-language reasoning for each decision.

3

Review Ranked Results

Trials are ranked as Strong Match, Possible Match, or Unlikely. See exactly which criteria were met, not met, or need verification.

Works with your AI provider

Choose the provider that fits your budget, privacy requirements, and infrastructure.

☁️ Anthropic (Claude)
☁️ OpenAI (GPT-4)
💻 Ollama (Local)
🔌 Any OpenAI-Compatible

Up and running in three commands

No API key needed — sandbox mode lets you explore the full interface with sample data immediately.

# Clone and install
git clone https://github.com/matthewhmaxwell/trialibre.git
cd trialibre/backend && pip install -e ".[dev]"

# Start the server (opens your browser)
trialibre serve

Open source. Free forever.

Trialibre is developed by the American Institute for Medical Research because clinical trial matching should be accessible to every clinician and researcher — not locked behind expensive proprietary platforms.

Beta: Trialibre is a screening tool, not a medical device. All matches must be verified by a qualified clinician. Handling real patient data requires a BAA with your chosen AI provider — see the privacy guide.