Molecular Intelligence for Any AI

Novo MCP

Your AI already knows chemistry. Pre-computed ADMET predictions, regulatory compliance, and research search across 122 million compounds — plus GPU docking, molecular dynamics, and end-to-end discovery pipelines. One connection to any MCP-compatible assistant. The platform handles the rest.

Get Started View setup

122M+Pre-computed compounds

84Properties each

<50msRetrieval

46Discovery tools

Quick setup

Any MCP-compatible AI client. Connect once — your AI gains molecular intelligence without new software.

Claude Desktop / Web / Mobile

1. Settings → Connectors → Add custom connector
2. Enter URL: https://ai.novomcp.com/mcp
3. Click Connect — a login page opens
4. Enter your API key
5. Done!

Claude Code (Terminal)

1. claude mcp add --transport http novomcp https://ai.novomcp.com/mcp
2. claude
3. /mcp  → authenticate in browser
4. Done! 46 tools now available

Cursor / Windsurf

1. Settings → MCP → Add server
2. Enter URL: https://ai.novomcp.com/mcp
3. Authenticate with your API key
4. Done! All tools available in chat

ChatGPT / Gemini / Copilot

1. Add MCP server in your client settings
2. Enter URL: https://ai.novomcp.com/mcp
3. Authenticate with your API key
4. Done! Molecular intelligence in your preferred AI

NemoClaw

1. Add NovoMCP as an MCP skill or tool
2. Server URL: https://ai.novomcp.com/mcp
3. Pass your API key via header: X-API-Key: nmcp_xxx
4. Done! 46 tools available to your agent

Other MCP Clients

Server URL: https://ai.novomcp.com/mcp
Transport: Streamable HTTP
Auth: OAuth 2.0 (automatic) or API key header

Then ask your AI: “What's the ADMET profile of aspirin?”

What your AI already knows

Start with the right target

108,000 target-disease associations ranked by a composite score — genetics, expression, and druggability. 56 pharmacogene profiles for patient stratification. 135,000 ClinVar resistance variants flagged for binding-site impact. Ask in plain language — “find targets for pancreatic cancer” — and get prioritized candidates with supporting evidence, ready to dock.

NovoMCP

Target Discovery — NSCLC

4 of 47 hits

108K

Target-disease pairs

Pharmacogenes

135K

Resistance variants

Ranked Targets

EGFR7SYD

Genetic · Expression · Druggable

0.94

score

KRAS6GJ8

Genetic · Druggable

0.89

score

MET4MXC

Expression · Druggable

0.81

score

ALK3LCT

Genetic

0.76

score

Scoring:genetics 0.40·expression 0.30·tractability 0.30

Already profiled. Already waiting.

84 properties per molecule — hepatotoxicity, cardiotoxicity, CYP450 interactions, drug-likeness, structural alerts. Pre-computed once across 122 million compounds. Retrieved in under 50 milliseconds. Novel compounds profiled on the fly with 31 ML models — without model queues or compute allocation.

ADMET Radar — Aspirin

CC(=O)Oc1ccccc1C(=O)O

85%

Abs

72%

Dis

91%

Met

78%

Exc

64%

Tox

Compliance as a compute layer

Eight jurisdictions — DEA, FDA, EPA, CWC, EU REACH, BTWC, Australia, OPCW — screened inline with every prediction. Controlled substance detection with scaffold pattern matching. Context-aware for intended use. Compliance runs inside the pipeline — without a manual checkpoint at the end.

FAVES Compliance — Aspirin

CC(=O)Oc1ccccc1C(=O)O

DEA

CLEAR

FDA

CLEAR

CWC

CLEAR

EPA

CLEAR

EU REACH

CLEAR

Overall Status

CLEAR

PAINS: 0

Structural Alerts: 0

Whitelisted FDA-approved drug

From lead to docked candidate, without hand-offs

Scaffold hopping generates structurally novel variants. Property-directed optimization targets specific profiles. GPU-accelerated docking scores candidates against protein targets. Every output compliance-screened before delivery — without switching between separate tools.

Terminal — optimize_molecule

Say a name. Get a structure.

Type “EGFR” or paste a sequence. The platform resolves experimental structures from RCSB or predicts new ones via OpenFold3 — with per-residue confidence scores — without navigating protein databases or managing compute infrastructure.

Structure Prediction — EGFR

OpenFold3

pLDDT

87.4

pTM

0.82

ipTM

0.79

Clash

2.1

Sequence · Confidence

MRPSGTAGAALLALLAALCPASRALEEKKVCQGTSNKLTQLGTFEDHFLSLQRMFNNCEVVLGNLEITYVQRNYDLSFLK...

Very High

High

Medium

Low

The complete landscape, one query

14,000 curated papers. 2,400 patents. 250,000 preprints. 500,000 clinical trials. 2.4 million bioactive compounds from ChEMBL. Semantic search that understands molecular context — without five separate searches across five separate databases.

Unified Research — Aspirin

5 sources · 1 query

Literature

Aspirin and cardiovascular disease prevention: a systematic review

Lancet · 2024 · Cited 142

94%

Patent

Novel aspirin formulations with improved bioavailability

USPTO · US2023/0142891 · Filed 2023

87%

Clinical Trial

ADAPTABLE: Aspirin Dosing in Cardiovascular Disease

Phase IV · Completed · n=15,076

91%

bioRxiv

COX-2 selectivity and platelet aggregation: mechanistic insights

bioRxiv · 2024 · doi:10.1101/2024.03.15

83%

ChEMBL

Bioactivity: COX-1 IC₅₀ = 1.67 µM, COX-2 IC₅₀ = 278 µM

CHEMBL25 · 1,247 assays · 892 targets

96%

Your warehouse, molecularly intelligent

Snowflake, Databricks, BigQuery, Supabase — pull compound libraries, run ADMET and compliance enrichment, push results back. Schema discovery and field mapping included — without ETL engineering or new infrastructure.

Terminal — pull_from_source

Binding, validated

GPU-accelerated AutoDock scores candidates against any protein target. Every pose checked against strain energy — so the ranking holds. Contact residues, binding distances, and delta affinities rendered in the chat — without opening a separate viewer.

NovoMCP

Docking Results (1PTH)

20 credits

PDB ID

1PTH

Resolution

1.8 Å

High Quality

Method

X-Ray

Binding Site

Known (co-crystal)

Binding Affinity (4 molecules)

#	SMILES	kcal/mol	Δ	Contacts
1	CC(=O)Oc1ccccc1C(=O)O	-9.4	—	4
2	CC(=O)Oc1ccc(O)cc1C(=O)O	-8.7	+0.7	5
3	CC(=O)Oc1cc(F)ccc1C(=O)O	-7.9	+1.5	3
4	c1ccc(NC(=O)C2CC2)cc1	-6.2	+3.2	2

Interactions — Best Binder (−9.4 kcal/mol)

TYR-385

H-bond (2.8 Å)

ARG-120

H-bond (3.1 Å)

VAL-349

Hydrophobic (3.9 Å)

LEU-352

Hydrophobic (4.1 Å)

Dynamics, not snapshots

GROMACS molecular dynamics at production scale — 100 ns trajectories with RMSD, RMSF, and hydrogen bond analysis. Jobs run async on GPU; results stream back with full trajectory plots. Your binding pose, validated across time — without managing a compute cluster.

NovoMCP

MD Simulation — COX-2 / Aspirin

Converged

Duration

100 ns

Temperature

300 K

Avg RMSD

1.8 Å

Binding ΔG

−8.2

kcal/mol

RMSD Trajectory

Backbone Cα

RMSF (avg)

0.9 Å

H-bonds

3.2 avg

Radius Gyration

22.4 Å

Long-running work, always visible

Docking, molecular dynamics, structure prediction — all run async on GPU. The jobs dashboard tracks every submission with live progress, ETAs, and auto-refresh. Keep working in chat while simulations run in the background — without polling or page reloads.

NovoMCP

Compute Jobs

Auto-refresh

Active

Completed today

Avg runtime

run_molecular_dynamicsRunning

md_7f3a8c

Started 8m ago67% · ~4m

dock_moleculesCompleted

dock_2e91b4

Started 12m ago100% · done

predict_structureRunning

struct_a4f2d1

Started 3m ago34% · ~6m

screen_libraryCompleted

screen_c8e5f9

Started 1h ago100% · done

Transparent by design

Credits consumed per tool, per request — visible in your chat any time you ask. No hidden costs, no surprise overages, no monthly reconciliation. Ask your AI “what have I used this month?” and see a complete breakdown — without logging into a separate billing portal.

NovoMCP

Credit Usage

Team Plan

Remaining Balance

9,113

of 10,000 monthly credits

Resets in

18 days

887 credits used this month (8.9%)

Top Tools This Month

dock_molecules340 · 17×

run_molecular_dynamics250 · 5×

predict_admet140 · 14×

search_literature95 · 19×

get_molecule_profile62 · 31×

See the results where you work

ADMET radar charts, 3D protein viewers, compliance dashboards, molecular structures — rendered inline in Claude, ChatGPT, and VS Code. Your team sees results and visualizations in the same conversation — without exporting to external tools.

Inline Visualizations

Rendered in your AI chat

ADMET Radar

FAVES Status

DEA

FDA

CWC

EPA

REACH

ALL

Properties

MW180.16

LogP1.31

QED0.55

3D Structure

Available tools

46 tools across 7 categories — all included with every account. Your AI selects and chains them automatically.

get_molecule_profileComplete molecular dossier — ADMET predictions, drug-likeness, regulatory status, and structural alerts. Pre-computed for 122M+ compounds, retrieved in under 50ms.

get_molecule_infoCore physicochemical properties via RDKit: molecular weight, LogP, TPSA, hydrogen bond donors/acceptors, rotatable bonds.

calculate_propertiesOn-demand property calculation for any SMILES: MW, LogP, LogD, TPSA, synthetic accessibility, Lipinski and Veber rule violations.

get_3d_properties32+ three-dimensional molecular descriptors from conformer generation: geometry, energy, electrostatics, surface area, and volume.

search_similarFind structurally related compounds using Tanimoto fingerprint similarity across 122M molecules. Configurable threshold.

filter_moleculesQuery the molecular database by property ranges, drug-likeness criteria, and compliance status. Exclude controlled substances automatically.

batch_profileProfile up to 100 compounds in a single request. Full ADMET, compliance, and property data for each.

target_discoveryIdentify drug targets from 108,000 genomic target-disease associations. Returns dockable targets ranked by evidence score with suggested PDB structures and druggability assessments.

validate_targetAdversarial target validation — stress-test a target hypothesis against clinical trials, ChEMBL bioactivity, and literature evidence. Returns tiered confidence score with supporting and contradicting evidence.

stratify_patientsPharmacogenomic patient stratification across 56 pharmacogene profiles. Predicts metabolizer phenotypes, dose adjustments, and population-level clinical viability.

predict_admet31 ML models predicting absorption, distribution, metabolism, excretion, and toxicity. Includes CYP450 substrate/inhibitor status, cardiotoxicity, hepatotoxicity, and nuclear receptor activity.

optimize_moleculeProperty-directed molecular optimization via NVIDIA MolMIM. Target specific QED, LogP, and similarity profiles. All outputs auto-screened through FAVES compliance.

check_complianceContext-aware regulatory assessment across eight jurisdictions. DEA, FDA, EPA, CWC, EU REACH, BTWC, Australia, OPCW. Controlled substance detection with scaffold pattern matching.

screen_libraryScreen compound libraries up to 1,000 molecules with full ADMET profiling and FAVES compliance assessment. Batch-optimized for throughput.

lead_optimizationScaffold hopping via RDKit substructure replacement. Generates structurally novel variants enriched with ADMET predictions and compliance scoring. Property-directed optimization available.

dock_moleculesGPU-accelerated molecular docking with AutoDock-GPU. Dock compounds against protein targets from PDB. Returns binding affinities, pose analysis, and interaction maps.

run_molecular_dynamicsGPU molecular dynamics with GROMACS. Simulate protein-ligand binding stability, conformational dynamics, and interaction energetics over nanosecond timescales.

audit_systemFree pre-flight protein classification. Detects membrane proteins, metal sites, heme/Fe-S clusters, and returns a routing verdict before committing GPU credits.

parameterize_metalQM-to-force-field bridge for metalloprotein MD via MCPB.py. Takes QM output, returns AMBER/GROMACS parameters with RESP charges for metal centers.

save_funnel_contextPersist discovery pipeline state — target, seed molecule, optimization results, docking scores — for session resumption across long-running discovery funnels.

get_funnel_contextResume a discovery pipeline from any saved checkpoint. Retrieve full context to continue from any step without re-running prior stages.

get_protein_structureSmart protein resolver. Accepts names ("EGFR"), PDB IDs, or sequences. Returns experimental structures from RCSB or de novo OpenFold3 predictions with per-residue confidence.

predict_structureDe novo 3D protein structure prediction via OpenFold3. Submit amino acid sequences, receive PDB files with confidence scores and visualization-ready coordinates.

get_structure_resultRetrieve completed structure predictions. Check job status, download PDB files, and access per-residue confidence metrics.

search_literatureSemantic search across 14,000+ curated drug discovery papers using vector embeddings. Returns titles, authors, abstracts, and DOIs ranked by relevance.

search_patentsSearch 2,400+ USPTO pharmaceutical patents by topic. Returns patent numbers, applicants, filing dates, and claim abstracts.

search_biorxivSearch 250,000+ bioRxiv and medRxiv preprints for the latest unpublished research. Full abstracts with DOIs.

search_chemblQuery 2.4 million bioactive compounds from ChEMBL. Returns SMILES, targets, IC50/EC50 bioactivity data, and assay metadata.

search_clinical_trialsSearch 500,000+ clinical trials from ClinicalTrials.gov. Filter by phase, status, condition, and intervention. Returns enrollment, sponsors, and timelines.

pull_from_sourcePull compound libraries from Snowflake, Databricks, BigQuery, or Supabase. Run ADMET and compliance enrichment. Push results back. Bidirectional with schema discovery.

push_to_destinationExport enriched molecular data to configured destinations. Schema preview, field mapping, and format conversion included.

get_platform_infoPlatform capabilities and statistics: database coverage, ADMET model inventory, compliance jurisdiction details, and API status.

get_job_statusTrack async jobs — structure predictions, MD simulations, batch processing. Returns status, progress, and results when complete.

get_credit_usageCredit balance, usage history, and cost breakdown by tool. Interactive dashboard with research value metrics.

Works with any MCP client

Claude. ChatGPT. Cursor. Windsurf. GitHub Copilot. VS Code. One URL, one API key, and every tool appears in your conversation — without SDKs to install, dashboards to learn, or context to switch.

Your workflow doesn't change. Your AI just thinks more clearly about chemistry.

One-click setup

Add the URL, authenticate once. OAuth handles the rest. New tools appear server-side — your setup never goes stale.

Natural language

Describe what you need. Your AI selects tools, chains them, and pursues the objective — pausing only when a decision needs your judgment.

Always current

New tools, new data, new models — deployed server-side. Your connection gains capabilities without reconfiguration.

Your AI already knows chemistry

One URL. One API key. 122 million compounds. Set up in sixty seconds.

Get Started