Molecule of the Week · May 13, 2026
PCL-RA-003 2D Structure (Thiophene-Piperazine JAK2 inhibitor)

PCL-PD-008

Novel Imidazopyridine-Pyrimidine KRAS G12D Inhibitor

O=C(c1cc(F)cc(F)c1)N1CCN(c2cn3cc(-c4ccnc(NC5CC5)n4)ccc3n2)CC1

Pancreatic Ductal Adenocarcinoma (PDAC) · KRAS G12D Inhibition

Selective KRAS G12D inhibitor with novel imidazopyridine-pyrimidine scaffold

Imidazo[1,2-a]pyridine core provides a flat heterocyclic scaffold for π-stacking in the KRAS Switch II pocket. Piperazine linker offers conformational flexibility and a basic nitrogen for salt-bridge interactions with Asp69. 3,5-Difluorobenzamide fills the hydrophobic allosteric pocket. Cyclopropylamino-pyrimidine head engages H95/D69 residues for G12D mutant selectivity over wild-type KRAS.

7.5

PROMISING — proceed to optimization

PharmaClaw Consensus Score (0–10)

MW

475.5

cLogP

3.61

QED

0.475

Tox Risk

LOW

SA Score

2.79

TPSA

78.7

Lipinski

Pass (0)

Agents Consulted

Chemistry Query Cheminformatics Pharmacology Toxicology Synthesis Catalyst Design Literature IP Expansion

Suggested Modifications

  • → Replace cyclopropylamine with azetidine for improved metabolic stability
  • → Add methyl to piperazine N for enhanced CNS penetration (brain metastases)
  • → Swap one fluorine for chlorine to modulate lipophilicity
  • → Introduce spirocyclic constraint on piperazine for rigidity and selectivity

Generated by PharmaClaw AI Pipeline v2.1.0 · Updated every Wednesday

View Archive →

For research purposes only. Not a therapeutic recommendation. All novel molecules require experimental validation.

PharmaClaw

AI-Orchestrated Drug Discovery Pipeline

Eleven domain-specialized agents integrate RDKit cheminformatics, PubChem data, FAERS safety signals, retrosynthetic planning, ADMET profiling, toxicology, IP analysis, and market intelligence — orchestrated by LangGraph on OpenClaw.

Live Demo Explore the Pipeline pip install pharmaclaw

Purpose-Built for Medicinal Chemistry

End-to-end drug discovery in one pipeline. From SMILES input to polished report — no black boxes, real-time data from PubChem, openFDA, and RDKit.

Chemistry & Retrosynthesis

RDKit property calculation, BRICS retrosynthesis, reaction feasibility scoring, PubChem integration, scaffold analysis, and MCS.

Cheminformatics & ADMET

3D conformers, pharmacophore mapping, 14 ML-trained ADMET models, Lipinski/Veber/QED rules, CYP/P-gp/hERG predictions.

Toxicology, IP & Market Intel

Structural alerts, PAINS/Brenk filters, FAERS signal detection, freedom-to-operate analysis, bioisostere suggestions, patent drafting.

Discovery Pipeline

⌨️

Input SMILES

↓
🧪

Chemistry

↓
🧬

Cheminformatics

↓
💊

Pharmacology

↓
☠️

Toxicology

↓
🔬

Synthesis

↓
💼

IP Check

↓
📊

Market Intel

↓
📄

Report

Each stage is a specialized AI agent. Chain them all in one command — or run individually.

Install in Three Commands

Ready for production use on your own infrastructure.

Step 1

pip install

pip install pharmaclaw[agents]
Step 2

Configure

pharmaclaw setup

Provide your OpenAI / Anthropic API key

Step 3

Run

pharmaclaw ask "Analyze EGFR T790M inhibitor"
Full Walkthrough → GitHub →

Pro Report Demo

Select a compound to generate a multi-agent chain report.

Select a compound above to generate a multi-agent Pro chain report.

Pro users get: Compound Comparison · PDF Export · Batch Mode · Synthesis Planning

See Pro Plans →

Pro Features

Unlock powerful workflows built for discovery teams.

Cheminformatics Agent

New

3D conformer ensembles (ETKDG + MMFF), pharmacophore mapping, RECAP fragmentation, stereoisomer enumeration, format conversion (SDF/MOL/PDB/XYZ/InChI).

Compound Comparison

Compare 2–5 SMILES side-by-side with ranked analysis across molecular properties, ADME, safety, and IP metrics.

PDF Report Export

Color-coded reports with 2D structure images, ADME tables, safety flags, and synthesis routes. One click to team-ready PDF.

Batch Mode

Upload a CSV of 1–500 compounds for parallel analysis. Aggregated results with per-compound scoring and CSV/PDF output.

Synthesis Planning

AI-powered retrosynthesis with multi-step route proposals, feasibility scoring, reagent availability, and estimated yields.

Enhanced ADME Profiling

80+ computed properties including CYP inhibition, BBB permeability, plasma protein binding, Pgp substrate prediction, and bioavailability.

Watch Lists & Alerts

Monitor drugs via FAERS safety signals with automated alerts. Coming soon: patent expiry watches and PubChem new-compound tracking.

Example: Compound Comparison Output

Property Compound A Compound B Compound C
MW315.7206.3560.6
LogP4.03.974.0
TPSA38.337.3102.0
HBD / HBA1 / 51 / 31 / 7
LipinskiPassPass1 violation
FAERS SignalsPsychiatric 22%GI bleed 28%Hepatotox 20%

Command-Line Interface

Nine agents. One terminal. JSON-native.

pip install pharmaclaw-cli

pharmaclaw

# Full pipeline with consensus scoring

$ pharmaclaw langgraph \

--smiles "CC1=NN(C(=O)C1=CC2=C(N=C(C=C2)NC3=CC(=NN3C)C(F)(F)F)Cl)C4CC4" \

--workflow full --verbose

LangGraph full pipeline starting...

[1/9] Chemistry — molecular properties...

[2/9] Cheminformatics — 3D conformers, pharmacophores...

[3/9] Pharmacology — ADME/PK profiling...

[4/9] Toxicology — safety profiling...

[5/9] Synthesis — route planning...

[6/9] Catalyst — recommendation...

[7/9] Literature — PubMed search...

[8/9] IP — freedom-to-operate...

[9/9] Market Intel — FAERS query...

Score: 7.5/10 — PROMISING — proceed to optimization

# Pipe agents together

$ pharmaclaw chemistry -s "CCO" | pharmaclaw toxicology | jq '.risk'

"Low"

Every Agent, One Command

Chemistry

pharmaclaw chemistry -s "CCO" --mode props

MW, LogP, TPSA, retrosynthesis, PubChem

Cheminformatics

pharmaclaw cheminformatics -s "CCO"

3D conformers, pharmacophores, RECAP, SDF/MOL/PDB

Pharmacology

pharmaclaw pharmacology -s "CCO"

Lipinski, Veber, QED, BBB, CYP3A4, P-gp, PAINS

Toxicology

pharmaclaw toxicology -s "CCO"

Safety profiling, structural alerts, risk scoring

Synthesis

pharmaclaw synthesis -s "CCO" --steps 3

Multi-step BRICS retrosynthesis, feasibility scoring

Catalyst Design

pharmaclaw catalyst --reaction suzuki

28 reaction types, Pd/Ru/Ni/Cu, novel ligand design

Literature

pharmaclaw literature -q "KRAS 2026"

PubMed + Semantic Scholar, TLDRs, citations

IP Check

pharmaclaw ip -s "CCO" --bioisosteres

FTO analysis, Tanimoto similarity, bioisosteres

Market Intel

pharmaclaw market --drug sotorasib

FDA FAERS adverse events, trends

LangGraph Multi-Agent Orchestration

4 Workflows

full — All 9 agents + conditional routing
~10s
quick — Chemistry + Toxicology
~2s
safety — Chem + Pharm + Tox
~4s
synthesis — Chem + Synth + Catalyst
~5s

Smart Features

Dynamic Routing — High tox? Auto-reroutes to Pharmacology
Self-Correction — SMILES validation, error recovery
Consensus Scoring — 0–10 with verdict and recommendations
Agent-Native — JSON in/out for any AI framework

Consensus Scoring

8–10
Excellent
6–7
Good
4–5
Fair
0–3
Poor

Compare

pharmaclaw compare -s "A,B,C"

Side-by-side multi-compound analysis

Batch

pharmaclaw batch -f compounds.csv

Process up to 500 compounds from CSV

Report

pharmaclaw report -s "CCO" -o out.json

Full pipeline → JSON report file

From Prompt to Novel Candidates in 3 Minutes

Input: "Create a novel lung cancer drug"

01

The Prompt

"Create a novel lung cancer drug"

One sentence. That's all it takes.

02

Pipeline Activates

Chemistry ✓ Cheminformatics ✓ Pharmacology ✓ Toxicology ✓ Synthesis ✓ Market Intel ✓ IP Analysis ✓ Report ✓
03

Real-World Safety Data

FDA FAERS Analysis — Osimertinib & EGFR Class

29,206 adverse event reports analyzed for osimertinib

6,134 diarrhoea reports across EGFR class — #1 safety concern

Rash (3,667 reports) — #2 unmet need

Live FDA FAERS data — not simulated, not cached.

04

3 Novel Compounds Designed

PharmaClaw-1

Chloroacetamide warhead (less reactive)

Lipinski: Pass SA: 2.79

Targets #2 adverse event (rash)

Top pick

PharmaClaw-2

N-methylpiperazine side chain

Lipinski: Pass Composite: 0.369 Feasibility: Easy

Targets #1 adverse event (diarrhoea)

PharmaClaw-3

Difluoromethoxy group

Lipinski: Borderline

Extended half-life for once-daily dosing

05

Head-to-Head Comparison

CompoundMWLogPQEDLipinski
Gefitinib446.94.280.518Pass
Erlotinib395.54.450.508Pass
Afatinib486.04.390.457Pass
PharmaClaw-2 ★497.64.260.369Pass
Osimertinib499.64.510.311Pass
06

Recommendation

PharmaClaw-2 recommended for advancement. Best balance of drug-likeness, synthetic feasibility, and rationale for reduced GI toxicity.

Next step: molecular docking against EGFR (PDB: 6JX0)

This entire analysis took 3 minutes and cost less than $0.04 in compute.

Try It Free See Pricing →

Pricing

Freemium to Enterprise.

Free

$0

  • Chemistry Query agent
  • Pharmacology profiling
  • Self-host on OpenClaw
  • Demo compound reports
Install Free
Most Popular

Pro

$29/mo founding member rate

$49/mo

  • All 9 agents + chaining
  • Compound comparison
  • PDF exports
  • Batch mode (500 runs/mo)
  • Synthesis planning
  • Enhanced ADME (80+ props)
  • Watch lists (10 watches)
  • API access
Start Pro Trial

Enterprise

Custom

  • Unlimited watches
  • Custom agents
  • On-prem deployment
  • Priority support
  • Unlimited runs
Contact Sales

FAQ

What is PharmaClaw? +

PharmaClaw is a suite of 11 AI agents built on OpenClaw for drug discovery. They chain together PubChem data, RDKit cheminformatics (2D and 3D), FAERS adverse event intel, ADME/PK profiling, toxicology screening, synthesis planning, catalyst design, literature search, and IP/patent analysis into unified workflows.

What input formats are accepted? +

SMILES strings, drug names, PubChem CIDs, and natural language queries. The Chemistry Query agent resolves names to structures automatically via PubChem.

Is PharmaClaw free? +

The Free tier includes Chemistry Query and Pharmacology agents. Pro ($49/mo, $29/mo founding members) unlocks all agents, multi-agent chaining, compound comparison, batch mode, synthesis planning, PDF exports, watch lists, and API access.

Can I self-host? +

Yes. PharmaClaw agents are OpenClaw skills. Install from ClawHub and run on your own infrastructure with your own LLM API key.

What is agent chaining? +

Agent chaining pipes the output of one agent into the next — Chemistry → Pharmacology → Toxicology → Synthesis → IP → Market Intel — all in one workflow. This is a Pro feature.

How accurate is the data? +

Agents query authoritative sources in real-time: PubChem, openFDA FAERS, and patent databases. RDKit computations are deterministic. AI-generated insights should be validated by domain experts before regulatory use.

What is synthesis planning? +

AI-powered retrosynthesis that breaks a target molecule into commercially available precursors, scores each route for feasibility, and suggests alternative pathways.

How do watch lists work? +

Watch lists monitor drugs for new FAERS safety signals and send alerts when significant changes are detected. Coming soon: patent expiry monitoring and PubChem new-compound watches.

Need Help?

We're here to help you get the most out of PharmaClaw.

Email Us

Questions, feedback, or partnership inquiries.

cheminem602@gmail.com

Report Issues

Bug reports, feature requests, and agent feedback.

GitHub Issues →

Community

Chat with other users and the OpenClaw community.

Join Discord →