# 💊 drug-pipeline-mcp

[![CI](https://github.com/DasClown/drug-pipeline-mcp/actions/workflows/ci.yml/badge.svg)](https://github.com/DasClown/drug-pipeline-mcp/actions/workflows/ci.yml)
[![Tests](https://img.shields.io/badge/tests-61_✔️-brightgreen)](https://github.com/DasClown/drug-pipeline-mcp/actions/workflows/ci.yml)
[![Docker](https://ghcr.io/dasclown/drug-pipeline-mcp/badge.svg?style=flat)](https://ghcr.io/dasclown/drug-pipeline-mcp)
[![Smithery](https://smithery.ai/badge/@crop-mcp/drug-pipeline)](https://smithery.ai/servers/crop-mcp/drug-pipeline)
[![Python 3.10+](https://img.shields.io/badge/python-3.10+-blue.svg)](https://www.python.org/downloads/)
[![License: MIT](https://img.shields.io/badge/license-MIT-green.svg)](LICENSE)
[![PyPI](https://img.shields.io/pypi/v/drug-pipeline-mcp?label=PyPI)](https://pypi.org/project/drug-pipeline-mcp/)
[![GitHub stars](https://img.shields.io/github/stars/DasClown/drug-pipeline-mcp?style=flat)](https://github.com/DasClown/drug-pipeline-mcp/stargazers)

**Pharmaceutical R&D Pipeline Intelligence for AI Agents** — a lightweight MCP server that aggregates clinical trial data, FDA/EMA approvals, safety surveillance (FAERS), drug labels, patents, drug interactions, recalls, and publications through a unified API. Every output includes a verifiable source identifier (NCT ID, FDA Application Number, or PMID).

Not a replacement for IQVIA/EvaluatePharma. A real-time, publicly accessible intelligence layer that complements subscription databases.

---

## Quick Start

```bash
pip install git+https://github.com/DasClown/drug-pipeline-mcp.git

# Start MCP server (stdio)
drug-pipeline

# Or HTTP mode for remote access
pip install drug-pipeline-mcp[http]
drug-pipeline --http --port 8081
```

Or deploy via [Smithery](https://smithery.ai/servers/crop-mcp/drug-pipeline) — one click, no config.

---

## Tools (28)

| Tool | What it does | Data Source |
|------|-------------|-------------|
| `search_trials` | Search clinical trials by condition, phase, status, sponsor | ClinicalTrials.gov |
| `get_trial_detail` | Full protocol for a specific NCT (eligibility, outcomes, locations) | ClinicalTrials.gov |
| `get_trial_results` | Trial outcomes — endpoints, adverse events, participant flow | ClinicalTrials.gov |
| `get_trial_sites` | Trial site locations — facilities, countries, geo distribution | ClinicalTrials.gov |
| `lookup_drug` | Drug info: active ingredients, strength, ATC classification, NDC | openFDA + RxNorm |
| `get_approvals` | FDA approval history with submission dates and status | openFDA Drugs@FDA |
| `get_eu_approvals` | EU/EMA authorization — brand names, ATC, status, orphan/biosimilar flags | EMA Daily XLSX |
| `get_safety_data` | FAERS adverse event reports — top reactions, serious outcomes, total count | openFDA FAERS |
| `detect_safety_signals` | Disproportionality screening (exploratory) — AE reporting ratios | openFDA FAERS |
| `get_drug_label` | FDA prescribing info — indications, boxed warnings, contraindications, dosing | openFDA Drug Labeling |
| `get_dailymed_label` | NIH drug label (OTC + Rx) — SPL set ID, version, DailyMed URL | DailyMed (NIH/NLM) |
| `get_recalls` | FDA drug recalls — Class I/II/III, reasons, dates, firms | openFDA Enforcement |
| `get_patent_expiry` | Patent & exclusivity — approval dates, market exclusivity estimates | openFDA |
| `get_drug_interactions` | Drug-drug interactions — FDA label + FAERS co-reported drugs | openFDA Labeling + FAERS |
| `get_drug_pricing` | US drug product ID — NDC codes, manufacturers, strength, form | openFDA NDC Directory |
| `get_opentargets_drug` | Drug-target MOA — mechanisms, targets, clinical stage, drug type | Open Targets (EMBL-EBI) |
| `get_us_orphan_designations` | US FDA Orphan Drug Designations — indication, status, exclusivity | MyChem.info |
| `approved_for_condition` | Find drugs by indication — which are EU-approved for a condition | EMA Daily XLSX |
| `list_orphan_drugs` | EU orphan drug designations — filter by therapeutic area | EMA Daily XLSX |
| `list_biosimilars` | EU biosimilars — filter by condition / therapeutic area | EMA Daily XLSX |
| `list_loss_of_exclusivity` | LOE timing — biosimilar competition by active substance | EMA + FDA |
| `company_pipeline` | Company R&D — trials grouped by phase + EU approval enrichment | ClinicalTrials.gov + EMA |
| `search_publications` | PubMed search for drug/trial publications | PubMed / NCBI |
| `find_investigators` | KOL / PI search — investigators by condition or drug | ClinicalTrials.gov + PubMed |
| `detect_combination_therapies` | Combination therapy detection — co-administered drugs in trials | ClinicalTrials.gov |
| `compare_drugs` | Head-to-head — FDA, EU, MOA, safety, patent across 2 drugs | Composite |
| `drug_pipeline` | **Composite** — drug info + FDA + EU + safety + label + signals + recalls + interactions + trials + pubs + patent | All sources |
| `pipeline_landscape` | **Full pipeline for a condition** — approved + Phase 3/2/1 + mechanisms + sponsors + pubs | Composite |


> ⚠️ **`detect_safety_signals`**: Exploratory disproportionality screening — uses an approximate denominator, not a validated PRR. Not for regulatory or clinical use. See methodology note in code.

---

## Architecture

```
drug-pipeline-mcp/
├── drug_pipeline/
│   ├── __init__.py        # Version
│   ├── server.py          # MCP server (28 tools)
│   └── sources.py         # Data source fetchers (API aggregation layer)
├── drug_pipeline_cli.py   # CLI entry point
├── tests/                 # 61 unit tests (pytest)
├── pyproject.toml
└── README.md
```

**Design philosophy:** Lightweight API aggregation. No caching layer. No ML models. No predictions. Each tool makes real-time requests to a public API and returns structured data with source identifiers. The server is intentionally simple — it extracts, structures, and annotates, nothing more.

**Limitations by design:**
- Rate limits apply per source (openFDA: 10 req/sec, ClinicalTrials.gov: generous)
- EMA data sourced from a daily-updated XLSX register — format changes monitored manually
- FAERS data is spontaneous reporting, not incidence rates
- The server does not interpret, predict, or synthesize beyond what the sources provide

---

## Data Sources

| Source | Data | Access |
|--------|------|--------|
| ClinicalTrials.gov | 500K+ studies, phases, status, eligibility, results, site locations | ✅ Always free |
| openFDA NDC Directory | Drug product ID, NDC codes, manufacturers | ✅ Always free |
| openFDA FAERS | Adverse event reports, reactions, serious outcomes | ✅ Always free |
| openFDA Drug Labeling | Prescribing info, interactions, contraindications | ✅ Always free |
| openFDA Drugs@FDA | Approval history, submissions, orphan designations | ✅ Always free |
| openFDA Enforcement | Recalls, market withdrawals, safety alerts | ✅ Always free |
| RxNorm / RxNav | Drug identifiers, RxCUI, ATC classification | ✅ Always free |
| PubMed / NCBI | Scientific publications, abstracts, PMIDs | ✅ Always free |
| EMA Medicines Register | EU authorization status, ATC, orphan/biosimilar flags | ✅ Always free |
| Open Targets (EMBL-EBI) | Drug-target mechanisms of action, clinical development stage | ✅ Always free |
| DailyMed (NIH/NLM) | Drug labels (OTC + Rx), structured product labeling | ✅ Always free |
| MyChem.info | US FDA Orphan Drug Designations | ✅ Always free |

**All sources are publicly funded and freely accessible.** No API keys, subscriptions, or licensing required.

---

## Verifiable Outputs

Every data point includes a direct link to its primary source:

| Output Field | Example Source URL |
|-------------|-------------------|
| NCT ID | `https://clinicaltrials.gov/study/NCT03178617` |
| FDA Application Number | `https://www.accessdata.fda.gov/scripts/cder/daf/index.cfm?event=overview.process&ApplNo=NDA125456` |
| PMID | `https://pubmed.ncbi.nlm.nih.gov/37272535/` |
| FDA Product NDC | `https://www.accessdata.fda.gov/scripts/cder/ndc/` |
| DailyMed SPL Set ID | `https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=xxx` |

No calculated fields. No predictions. No estimates. The tool is an aggregator, not an oracle — it brings primary-source data into an AI agent's context so the LLM can apply reasoning, not so the server can produce answers.

---

## Testing & Quality

| Check | Status |
|-------|--------|
| Unit tests | ✅ 61 passing (pytest) |
| CI/CD | ✅ Multi-Python matrix (3.10–3.13), Docker build, PyPI publish |
| Linting | ✅ Ruff (zero warnings) |
| Formatting | ✅ Black-compatible |
| Code coverage | Tracked in CI |

---

## Regulatory Intelligence

Beyond drug-level approvals, this project provides **multi-jurisdiction regulatory framework intelligence** for pipeline analysis. See [`docs/local-regulation-2026.md`](docs/local-regulation-2026.md) for a comprehensive reference covering 7 jurisdictions:

- 🇺🇸 **US IRA** — Medicare Part D price negotiation (Sep 2026), Small Molecule Penalty (7 yr vs 13 yr)
- 🇩🇪 **Germany AMNOG** — Benefit assessment, 2026 reform with fixed effect-size thresholds
- 🇫🇷 **France HAS/CEPS** — SMR/ASMR ratings, 400–600 day access timelines
- 🇮🇹 **Italy AIFA** — 21 regional formularies, payback mechanisms
- 🇬🇧 **UK MHRA/NICE** — Post-Brexit ILAP pathway, £/QALY thresholds
- 🇯🇵 **Japan PMDA/NHI** — Sakigake designation, biennial price revision
- 🇨🇳 **China NMPA/NRDL** — Annual –61% price negotiation

---

## Example Agent Queries

> *"What's in the pipeline for GLP-1 agonists?"*
→ `drug_pipeline(drug_name="semaglutide")` → ATC class, FDA status, clinical trials, publications

> *"Which companies have Phase 3 trials for non-small cell lung cancer?"*
→ `search_trials(condition="non-small cell lung cancer", phase="PHASE3", status="RECRUITING")`

> *"Is pembrolizumab approved in the EU vs US?"*
→ `get_approvals(drug_name="Keytruda")` + `get_eu_approvals(drug_name="Keytruda")`

> *"What are the safety signals for semaglutide?"*
→ `get_safety_data(drug_name="semaglutide")` + `detect_safety_signals(drug_name="semaglutide")`

> *"What does the label say for Keytruda?"*
→ `get_drug_label(drug_name="Keytruda")` → indications, boxed warnings, contraindications, dosing

> *"When does the patent for Keytruda expire?"*
→ `get_patent_expiry(drug_name="Keytruda")` → exclusivity information

> *"What drugs are approved for non-small cell lung cancer in the EU?"*
→ `approved_for_condition(condition="non-small cell lung cancer")`

---

## Client Integration

### Claude Desktop

```json
{
  "mcpServers": {
    "drug-pipeline": {
      "command": "python3",
      "args": ["-m", "drug_pipeline.server"]
    }
  }
}
```

### Cursor / VS Code

```json
{
  "mcpServers": {
    "drug-pipeline": {
      "command": "uvx",
      "args": ["drug-pipeline-mcp"]
    }
  }
}
```

### HTTP / SSE (Remote)

```bash
pip install drug-pipeline-mcp[http]
drug-pipeline --http --port 8081
```

Connect at `http://your-server:8081/sse`.

### Smithery

[One-click deploy](https://smithery.ai/servers/crop-mcp/drug-pipeline). No config needed.

---

## 🤝 Getting Help & Contributing

| Channel | Purpose |
|---------|---------|
| **[💬 GitHub Discussions](https://github.com/DasClown/drug-pipeline-mcp/discussions)** | Questions before coding, feature ideas, community chat |
| **[🐛 GitHub Issues](https://github.com/DasClown/drug-pipeline-mcp/issues)** | Bug reports, confirmed feature requests |
| **[📖 CONTRIBUTING.md](CONTRIBUTING.md)** | Development setup, code style, testing |

New contributors welcome. See [CONTRIBUTING.md](CONTRIBUTING.md) for setup instructions.

---

## License

MIT