NovoMCP

Showing 25 of 41 from the live probe.

• Get Molecule Profile

  Full molecular profile with ADMET predictions and regulatory compliance. This is the PRIMARY tool for profiling any molecule. For known molecules (in 122M database): returns pre-computed ADMET + FAVES compliance. For other molecules: automatically runs ML-based ADMET predictions + compliance checks. Always returns complete data — no follow-up tools needed.

• Get Molecule Info

  Quick lookup of basic molecular properties only (MW, formula, LogP, TPSA, H-bond counts). Lightweight — no ADMET or compliance data. Use get_molecule_profile instead if you need a complete analysis.

• Get Platform Info

  Get NovoMCP platform information including subscription tiers, available tools per tier, database statistics, ADMET capabilities, compliance lists, and credit usage. Use info_type='usage' to see your organization's credit balance and consumption.

• Push to Destination

  Push tool results to a connected destination such as Google Sheets or BigQuery.

• Pull from Source

  Pull compound data from a connected data warehouse (Snowflake, Databricks), run ADMET/compliance/optimization, and optionally push enriched results back. Actions: preview (inspect table), pull (read rows), estimate_pipeline (cost estimate), execute_pipeline (run full pipeline).

• Get Credit Usage

  Check your NovoMCP credit balance and research value realized. Shows included credits, overage costs, and billing period. 1 credit = $1. Use when users ask 'How many credits?', 'What's my usage?', 'Summarize my spend', or 'Check my account'.

• Search Similar Molecules

  Find structurally similar molecules in the 122M compound database. Returns similar molecules with their pre-computed ADMET properties and FAVES compliance status.

• Filter Molecules

  Filter the 122M molecule database by properties. Returns molecules matching criteria with pre-computed ADMET and FAVES data.

• Batch Profile Molecules

  Batch version of get_molecule_profile: ADMET (toxicity incl. hepatotoxicity, CYP metabolism, nuclear receptors, stress response), FAVES compliance, and properties for up to 100 molecules in one call. Pre-computed for known molecules; novel ones get on-the-fly properties + ML ADMET. Set include_admet=false for faster, cheaper properties-only screening.

• Optimize Molecule

  Property-directed molecular optimization using NVIDIA MolMIM (generative AI). Given a seed molecule and target objectives (QED, LogP, TPSA, similarity), generates structurally similar variants biased toward the desired property profile. Returns 3-10 optimized SMILES with property deltas vs seed, each auto-checked for FAVES compliance. Keeps structural similarity high (Tanimoto > 0.4 typical) — for diverse scaffold hopping, use lead_optimization instead.

• Generate File Upload URL

  Generate a signed upload URL for large files (QM logs, PDB structures, compound libraries, trajectories). Upload the file directly to the URL — no data flows through the chat. Returns a file_id to reference in downstream tool calls. Free (0 credits). Upload URLs expire in 30 minutes. If auto_process is set, the file will be processed automatically after upload — no need to call a second tool.

• Get File Status

  Check the status of an uploaded file — pending, uploaded, processing, completed. Shows linked tool calls, job IDs, and processing results. Use after uploading a file to verify it's ready for tool consumption.

• List Uploaded Files

  List files uploaded to your account. Filter by file type (qm_log, pdb, trajectory, library) or status (uploaded, processing, completed). Shows file IDs, names, sizes, and linked jobs.

• Get Pipeline Audit Log

  Retrieve the per-molecule audit trail for a completed pipeline execution. Shows what happened to each molecule: input SMILES, standardization, per-tool results (ADMET, compliance, properties), disposition (included/excluded), and exclusion reasons. Use for GxP compliance documentation.

• 'Novo AG' / '/agm' → Autonomous Mode Trigger

  TRIGGER KEYWORDS: 'Novo AG', 'novo ag', 'novoag', '/agm', '/AGM' (case-insensitive). Call ONLY when the user's message contains one of these literal tokens. Returns the 12-stage autonomous discovery instruction set — read it and follow it. For messages without these keywords, operate in normal collaborative mode (use tools as needed, ask clarifying questions). Pass mode='interactive' for the human-in-the-loop variant. Cost: 0 credits.

• Save Funnel Stage

  Record a HUMAN-REVIEWED checkpoint decision in the discovery funnel. **You do NOT need to call this to log tool calls — every tool call is already auto-logged server-side as an 'exploration' event under the session funnel_id. NEVER ask the user whether to log; logging is automatic and client-agnostic.** Call save_funnel_stage ONLY to capture the user's explicit decision/approval at a reviewed checkpoint in an interactive funnel (pass human_reviewed: true with human_decision + human_prompt) — tha

• Get Funnel Audit Log

  Retrieve the reproducibility log for a discovery funnel run. Shows every event (checkpoints + exploration tool calls) with tool arguments, result summaries, AI recommendations, human decisions, molecule counts, filtering details, and compute costs. Filter by event_type='checkpoint' for a clean peer-review view that excludes ad-hoc exploration.

• List Discovery Funnels

  List recent discovery funnel runs with metadata — disease, target gene, outcome, stage count, credits consumed, best affinity. Use this to find a funnel_id before calling get_funnel_audit. Returns the most recent funnels for the current org, enriched with terminal summary data when available.

• Save Funnel Memory (terminal summary)

  Write a terminal summary of a completed discovery funnel to cross-run memory. Called once at the end of a funnel run (typically after Stage 12 completion). Captures target metadata, outcome, failure patterns, key decisions, and a natural-language summary for analogical retrieval. Powers search_prior_runs — future funnels targeting the same gene/disease can learn from this run's outcome and avoid repeating mistakes. Semantic embedding generated automatically from the summary via Azure OpenAI text

• Search Prior Discovery Runs

  Query cross-run memory for past discovery funnels that targeted the same gene, PDB, or therapeutic area. Returns terminal summaries, outcomes, and lessons from prior attempts. Call at funnel start to learn from precedents — avoid repeating known failure modes, reuse successful scaffolds, calibrate threshold expectations. Includes a lazy backstop that auto-generates template summaries for completed funnels that lack explicit memory entries, ensuring cross-run coverage is complete.

• Calculate Properties

  Calculate RDKit molecular properties on-demand. Returns Lipinski descriptors, drug-likeness scores (QED, SA_Score), physicochemical properties (LogP, TPSA, MW), and structural features. Does NOT include ADMET or compliance — use get_molecule_profile for a complete analysis.

• Predict ADMET

  Predict toxicity and ADMET properties: cardiotoxicity, hepatotoxicity, nephrotoxicity, carcinogenicity, CYP450 inhibition (1A2/2C9/2C19/2D6/3A4 substrate + inhibitor), nuclear receptor activity (AR/ER/PR/GR/PPAR), stress response (p53, oxidative stress), absorption, distribution, metabolism, excretion. Returns per-model probabilities with severity categories. 40+ ML models from addie-models backend. Normally called automatically by get_molecule_profile; use directly for ADMET-only queries.

• Predict Clinical Outcomes

  Predict Phase I clinical trial clearance probability for a small molecule. Automatically gathers all 63 required features by orchestrating chem-props (physicochemical), faves-compliance (structural alerts, BOILED-Egg), and addie-models (ADMET) in parallel, then calls the NovoExpert v3 model. Returns a calibrated probability, SHAP feature explanations, and a domain-specific competence assessment. The model is validated for CARDIOVASCULAR and mainstream compounds (AUROC 0.72-0.76) but NOT for onco

• Search Literature

  Find published journal articles and research papers on drug discovery topics. Searches 14,398 curated peer-reviewed publications via Pinecone semantic search. Returns paper titles, abstracts, authors, DOIs, and relevance scores. Covers ADMET research, target validation, medicinal chemistry, SAR studies, and clinical pharmacology. Use for literature review, prior art assessment, and evidence gathering during target evaluation.

• Search Patents

  Search granted and pending USPTO pharmaceutical patent filings (2,416 documents, Pinecone semantic search). Returns patent titles, abstracts, applicants, filing dates, and relevance scores. Use for intellectual property landscape analysis, freedom-to-operate assessment, prior art search, and competitive intelligence on patented scaffolds or mechanisms.