MCP tool suite PV / CSW
An independent Provotics MCP reference implementation tested for use with Claude. Twenty-two local handlers are implemented in the research codebase, but production entitlement integration and a Provotics-hosted MCP endpoint are not deployed. This catalog is a research preview, not a currently purchasable service.
The site-of-origin handler exposes the Provenance-2 flagship core and its uncertainty controls. Other handlers are exploratory research integrations with narrower evidence, assay-confirmation caveats, or unresolved release gates. None is a clinical tool.
Deployed core · local integration
the engine the suite calls
The flagship cancer site-of-origin core and its trust layer. The local reference implementation can call it across 25 anatomical sites with calibrated confidence and abstention; that does not make the Workbench itself a deployed hosted service.
Implemented reference handler
classify_tumor_origin
Reads a tumor's site of origin from RNA-seq and returns calibrated confidence, the top alternatives, and an abstain option when no site is clearly ahead.
Implemented reference handler
trust_envelope
The honesty layer in one call: a calibrated probability, a conformal candidate set, a novelty flag, and an explicit abstain decision, so an agent can reason about whether to trust the call.
Implemented reference handler
novelty_check
Flags a profile that sits off the training manifold, so the model declines inputs it should not answer instead of guessing.
Exploratory reference handler
explain_prediction
Surfaces the genes that drove a specific prediction, so the call stays reviewable rather than a black box.
Implemented reference handler
panel_coverage_report
Reports how well your genes cover the model's panel, high, partial, or low, without ever revealing the panel contents.
Implemented reference handler
model_card_facts
Returns frozen, cited model-card facts, so an agent grounds its answers in real numbers and never invents a Provotics figure.
Implemented reference handler
harmonize_expression
Confirms a profile is accepted and mapped into the model's space, so inputs from different pipelines are comparable. It never returns your per-gene values back.
Exploratory reference handler
research_report
A deterministic, research-use-only report composed from the model's own outputs, with no language model in the loop and nothing invented.
Exploratory reference handler
proliferation_index
A within-type proliferation percentile, reported as a coherence signal, never a validated prognostic score.
Exploratory reference handler
immune_microenvironment
The tumor's immune phenotype, from cold to inflamed, as a descriptive research signal.
Exploratory reference handler
cohort_harmonize_classify
Batch cohort-alignment classification for a single-site study of 20 or more samples, never a single tumor.
Exploratory reference handler
molecular_portrait
Bundles the site call with separately labeled exploratory heads. It is not a single validated whole-tumor portrait; each field retains its own evidence scope and caveat.
Exploratory reference handler
msi_status
An expression-based microsatellite-instability signal, site-aware and reported research-grade. Confirm on a PCR or IHC assay; not a diagnostic call.
Exploratory reference handler
molecular_subtype
A molecular-subtype signal within the predicted site, where a subtype head exists (for example PAM50 on breast). Research use only.
Exploratory reference handler
driver_alterations
Expression-imputed candidate driver alterations, surfaced as research hypotheses. Confirm on a DNA assay before any use; never a call.
Exploratory research handler
drug_pathway_sensitivity
A within-patient research signal of pathway sensitivity. Not a therapy recommendation, and the underlying data license is under review.
Exploratory research handler
chembl_compound_lookup
Looks up compound and target context from ChEMBL for this tumor's over-expressed targets, as research leads, never treatment recommendations.
Exploratory reference handler
binding_affinity
An in-silico affinity estimate for a molecule against a target's commercial-clean model, shown only inside the model's applicability domain. Out-of-domain molecules and targets without a clean model abstain. Not measured binding; confirm in a wet lab.
Exploratory reference handler
admet_liability
A public-domain liability panel for a molecule (hERG, CYP, the Tox21 panel, mutagenicity, hepatotoxicity). Each is a risk probability, not a safety pass; a model below its bar abstains. Nothing is ever labelled safe or clear.
Exploratory research handler
design_candidate
Generates novel candidate molecules for a target with in-silico scores, with potency shown only inside a commercial-clean model's domain and every candidate screened for hazardous motifs first. A candidate is a research hypothesis for a wet lab, never a drug.
Exploratory research handler
druggable_target_scan
Maps a tumor's imputed drivers to the druggable targets the engine can design against with a defensible, commercial-clean model. Today one target qualifies; the rest are shown honestly as dark, not implied.
Exploratory research handler
methylation_class
A DNA-methylation tumor-class signal with a calibrated 90% prediction set, an off-array platform-shift warning, and abstention. A separate modality from the RNA-seq classifier that, on early cross-array evidence, is more stable across array platforms. Confirm with an orthogonal assay.
Forthcoming
tumor_purity_estimate
Tumor purity estimation. Not yet live: it returns an honest forthcoming notice until the clinical purity labels land.
Forthcoming
multiomics_integrate
Paired multi-omics integration. Not yet live: pending paired data matrices on the same cases, and returns a forthcoming notice until then.
Forthcoming
depmap_drug_response
A cell-line drug-response signal. Not yet live: the DepMap source needs a per-dataset licensing review before any build, so it returns an honest forthcoming notice. Research use only, never a therapy recommendation.
Twenty-two handlers pass the local reference test suite; three remain forthcoming and return an explicit unavailable notice. A passing local handler test establishes implementation behavior, not scientific validation or production availability. Hosting, live entitlement verification, operational quotas, and legal clearance for license-restricted data must be completed before customer access opens.
Access
No public MCP endpoint or paid Workbench access is available today. Register interest for a technical preview; access will open only after hosting, entitlement, security, and licensing gates are complete.