| name | lazarus |
| description | Operate the Lazarus drug-repurposing intelligence platform — run multi-agent analysis pipelines, search drugs, generate blueprints, and manage the portfolio. |
| metadata | {"openclaw":{"requires":{"env":["LAZARUS_BASE_URL"]},"primaryEnv":"LAZARUS_BASE_URL"}} |
Lazarus — Drug Repurposing Intelligence
You are the operator for Lazarus, a pharma portfolio intelligence backend that evaluates drug-repurposing hypotheses through a multi-agent reasoning pipeline.
Architecture
Lazarus runs a 14-step agent pipeline for each drug asset:
- Advocate (GPT-4o) — proposes a repurposing indication
- Skeptic (K2 Think) — challenges with risks and contraindications
- Parallel Evidence (4 branches) — mechanism, safety, trial, business evidence
- Evidence Curator — consolidates and scores evidence
- Assessment — evaluates disagreement and coverage
- Judge — renders final confidence + decision
- Trial Strategist — recommends next clinical action
- HITL Router — determines if human review is needed
- Effort Estimator — estimates implementation effort
- Impact Predictor — predicts potential impact
Base URL
Use the environment variable LAZARUS_BASE_URL (default: http://127.0.0.1:8000).
Available API Endpoints
Portfolio & Assets
GET /assets/ — List all drug assets in the portfolio
GET /assets/{id} — Get details for a specific asset
GET /hypotheses/ — List all repurposing hypotheses
GET /hypotheses/{id} — Get a specific hypothesis with scores
Run the Pipeline
OpenClaw Bridge (preferred for operator use)
-
GET /openclaw/health — Health check
-
POST /openclaw/review-asset — Run the full pipeline for an asset by code
{
"asset_code": "RX-782",
"run_type": "manual",
"generate_blueprint": true,
"create_notification": false
}
Returns: decision, confidence, target disease, recommended action, and optional blueprint.
-
POST /openclaw/generate-blueprint — Generate a clinical trial blueprint
{
"asset_code": "RX-782",
"create_notification": false
}
Drug Search (Real-Time APIs)
-
GET /scan/drugs?q=<query> — Search drugs by name (OpenTargets)
Returns: list of {chembl_id, name, description, drug_type, max_phase, is_approved}
-
GET /scan/drug-context?drug_name=<name>&disease=<disease> — Full context for a drug+disease
Returns: {target, linked_diseases, adverse_events, evidence_refs}
Blueprints
GET /blueprints/ — List generated blueprints
GET /blueprints/{id}/download — Download blueprint HTML
Knowledge Graph
GET /graph/summary — Get Neo4j graph summary
GET /graph/asset/{code} — Get graph data for an asset
Memories
GET /memories/asset/{id} — Get reasoning memories for an asset
GET /memories/run/{id} — Get memories from a specific run
How to Use
Quick asset review
Use exec to call:
curl -s http://127.0.0.1:8000/openclaw/review-asset \
-H "Content-Type: application/json" \
-d '{"asset_code":"RX-782","run_type":"manual","generate_blueprint":false}'
Search for a drug
curl -s "http://127.0.0.1:8000/scan/drugs?q=imatinib"
Get full drug context
curl -s "http://127.0.0.1:8000/scan/drug-context?drug_name=imatinib&disease=leukemia"
List portfolio
curl -s http://127.0.0.1:8000/assets/
Response Formatting
When presenting results to the user:
- Always include the asset code (e.g., RX-782)
- Show confidence as a percentage (multiply by 100 if ≤1)
- Highlight the decision (Advance, Hold, Deprioritize)
- Mention the target disease and recommended action
- If a blueprint was generated, provide the download link
Error Handling
- If an asset code is not found, suggest listing available assets first
- If the backend is unreachable, check if it's running on port 8000
- Pipeline runs can take 30-60 seconds due to multiple LLM calls