MCP tool reference

The Nullary MCP server exposes 31 tools. Install it first (guide), then your assistant can call any of these.

Common inputs & response

Every search_* tool accepts these optional inputs (plus any tool-specific ones noted below):

Param	Type	Description
`target`	string	UniProt accession or gene symbol (fastest with an accession).
`compound`	string	Compound name, ChEMBL ID, or InChIKey.
`indication`	string	Disease / indication (where applicable).
`outcome`	string	`inactive`, `terminated`, `failed_safety`, …
`limit` / `offset`	integer	Pagination. `limit` 1–100 (default 25).

Each tool returns a stable envelope. Results default to the curated tier (the credible, dose-response / human-curated layer); every finding carries full provenance (source, DOI/PMID/URL, license, confidence).

json

{
  "findings": [ /* finding objects (same shape as REST /findings) */ ],
  "returned_count": 12,
  "has_more": true,
  "query_metadata": { "tool": "search_inactive_compounds", "limit": 25, "offset": 0 },
  "coverage_note": "…"   // present only for thin Phase-1 modalities
}

This envelope is also returned as MCP structured output: every tool declares an outputSchema, and tools/call returns the same object as structuredContent alongside the text block — so structure-aware clients get a typed result without parsing JSON out of a string. All tools are read-only (annotated readOnlyHint).

Small molecule

Tool	Returns
`search_inactive_compounds`	Inactive compound–target pairs.
`search_failed_selectivity`	Small molecules that failed selectivity.
`search_admet_failures`	Small-molecule ADMET failures (hepatotox, cardiotox, oral bioavailability, BBB, …).

CRISPR

Tool	Returns
`search_failed_guides`	Failed / ineffective guide RNAs.
`search_failed_essentiality_screens`	Non-dependency / failed essentiality screens.
`search_ancestry_specific_failures`	Ancestry-specific CRISPR failures.

Antibody

Tool	Returns
`search_developability_failures`	Developability failures: aggregation, polyreactivity, expression, thermostability.
`search_failed_clinical_antibodies`	Discontinued / terminated clinical antibodies.

Peptide

Tool	Returns
`search_failed_peptide_therapeutics`	Failed peptide therapeutics.
`search_peptide_stability_issues`	Peptide stability / half-life failures.

PROTAC

Tool	Returns
`search_failed_protacs`	PROTACs that failed degradation / ternary / permeability.
`search_protac_e3_issues`	E3-ligase recruitment / ternary-complex failures.

Oligonucleotide

Tool	Returns
`search_failed_oligonucleotides`	ASOs / siRNAs that failed engagement or developability.
`search_oligo_delivery_failures`	Oligonucleotide delivery failures.

Vaccine, ADC & bispecific

Tool	Returns
`search_failed_vaccines`	Failed / terminated vaccines (by pathogen or indication).
`search_vaccine_immunogenicity_failures`	Failed vaccine immunogen designs.
`search_failed_adcs`	ADCs that failed at any stage.
`search_adc_linker_failures`	ADC failures attributed to linker chemistry.
`search_failed_bispecifics`	Bispecifics that failed at any stage.
`search_bispecific_format_failures`	Bispecific format / chain-pairing failures.

Cross-cutting

Tool	Returns
`search_admet_failures_all_modalities`	ADMET failures across every modality.
`search_drug_drug_interaction_failures`	Drug–drug interaction failures (PK/PD).
`search_mechanism_failures`	Mechanism-of-action failures.
`search_failed_replications`	Findings that failed to replicate.
`search_safety_failures`	Clinical / preclinical safety failures across modalities.

Cross-modality history

These answer “what's been tried” across every modality. Each takes one required argument.

Tool	Returns
`search_target_history`	All failed approaches against a target. Requires target (UniProt accession is fastest).
`search_indication_history`	All failed approaches for an indication. Requires indication.
`search_pathogen_history`	Vaccine / antimicrobial failures for a pathogen. Requires pathogen.

Direct lookups

Tool	Returns
`get_compound`	A molecule's full negative profile across sources. One of inchi_key, chembl_id, or pubchem_cid.
`get_finding_provenance`	A single finding with full provenance. Requires id (finding UUID).
`get_coverage`	Per-modality and per-source coverage counts. No arguments.

Example: get_compound

json

// tools/call → get_compound
{ "chembl_id": "CHEMBL941" }

// →
{
  "compound": { "pref_name": "IMATINIB", "chembl_id": "CHEMBL941",
                "inchi_key": "KTUFNOKKBVMGRW-UHFFFAOYSA-N", "max_phase": 4 },
  "returned_count": 200,
  "findings": [ /* every negative finding for this molecule, across sources */ ]
}

Prefer raw HTTP? The same data is available without MCP — see the REST API reference.