BioAgentics

Multi-agent system for biomedical research. BioAgentics uses coordinated AI agents to identify research opportunities, build computational tools, analyze data, and advance our understanding of disease biology. Research is organized into divisions — independent research domains with their own agent role definitions, plans, and output.

Live dashboard: bioagentics.mtingers.com

Research Divisions

Division	Focus	Key Data Sources
Cancer	Genomic analysis, drug discovery, biomarkers, treatment resistance	TCGA, DepMap, COSMIC, PRISM, ChEMBL
Crohn's Disease	Microbiome, mucosal immunology, IBD therapeutics	IBDGC, HMP, RISK cohort, MetaHIT
Tourette Syndrome	CSTC circuits, neuroimaging, tic disorder genetics	TSAICG, ENIGMA, ABCD Study, EMTICS
PANDAS/PANS	Autoimmune neuropsychiatry, molecular mimicry, anti-neuronal antibodies	ImmPort, IEDB, GAS genomics, Cunningham Panel
Diagnostics	Making diagnosis more accurate, accessible, and affordable — any disease	TCIA, PhysioNet, ISIC, Grand Challenges, UK Biobank

Each division has its own role definitions in org-roles/{division}/, research plans in plans/{division}/, and output in output/{division}/.

Architecture

BioAgentics is a single system with specialized agents that coordinate through a shared API. Each division runs its own set of agents independently.

Agent	Role
Research Director	Identifies research opportunities, designs studies, directs scientific strategy
Literature Reviewer	Scans for relevant publications, methods, datasets, and new opportunities
Data Curator	Manages datasets, verifies data sources, organizes the data directory
Project Manager	Coordinates research initiatives from plan to completion
Developer	Implements data pipelines, analysis tools, and computational models
Analyst	Runs analyses, interprets results, flags novel and promising findings
Validation Scientist	Validates scientific rigor, code correctness, and reproducibility
Research Writer	Documents methodology, findings, and maintains the knowledge base
Systems Engineer	Improves the BioAgentics system itself — codebase, tooling, agent configs

Research Pipeline

Research Director → Project Manager → Developer → Analyst → Validation Scientist → Research Writer
     (propose)        (plan tasks)     (build)    (analyze)     (validate)          (document)

Supporting agents run continuously:

• Literature Reviewer feeds new papers and methods to the Research Director
• Data Curator monitors data sources and organizes datasets
• Systems Engineer improves the platform itself

Labeling System

Research initiatives are tagged with labels for tracking significance:

• drug-candidate, drug-repurposing — therapeutic potential identified
• novel-finding — unexpected or previously unreported result
• biomarker — diagnostic/prognostic marker candidate
• high-priority — results warrant urgent follow-up
• promising — early positive signals
• cost-reduction, accessibility — cheaper or more accessible diagnostic/therapeutic approaches
• ai-diagnostic, point-of-care, screening, imaging — diagnostic modality
• Domain-specific: genomic, microbiome, immunology, neuroimaging, autoimmune, comorbidity, clinical, multi-omics

Getting Started

Prerequisites

• Python 3.13+
• uv package manager
• Claude Code CLI

Setup

# Install dependencies
uv sync

# Start the local coordination API
uv run python -m bioagentics.agent_api.main

# Start the dispatcher (in another terminal)
uv run python -m bioagentics.dispatch

Configuration

• agents.toml — divisions, agent roles, dispatch timing, and model settings
• .env — API URL and API key

Project Structure

src/bioagentics/          # All code lives here
  config.py               # Configuration and API client
  dispatch.py             # Agent lifecycle and scheduling
  mcp_server.py           # MCP tools for agent coordination
  agent_api/              # FastAPI coordination server + web UI
org-roles/                # Agent role definitions
  cancer/                 # Cancer division roles
  crohns/                 # Crohn's disease roles
  tourettes/              # Tourette syndrome roles
  pandas_pans/            # PANDAS/PANS roles
  diagnostics/            # Diagnostics roles
plans/{division}/         # Research initiative plans (created by Research Director)
output/{division}/        # Research output artifacts (data, figures, reports)
data/                     # Research data, datasets, results (created at runtime)
cache/                    # Agent context summaries (managed by dispatcher)
agents.toml               # Agent and division configuration