Skip to main content
Glama

mosaic-mcp

Coverage-honest pre-clinical target intelligence over MCP. Every count carries a coverage state, so "we did not look" never reads as zero.

3.0.0 is a thin client to the hosted Mosaic service. It serves Mosaic's five MCP tools over stdio and forwards every call to https://mcp.getmosaic.dev/mcp with your API key. It holds no database and no query layer: the answers, and any refusal, are the hosted service's.

Deprecated: the 2.x bring-your-own-database mode. Versions up to 2.2.1 ran Mosaic's query layer over a PostgreSQL database you provided (DATABASE_URL). That mode is deprecated and is not in 3.0.0; DATABASE_URL is ignored. Its eleven tools answer an old client by name with what replaced them. To keep running it, pin mosaic-mcp<3 (2.2.1 is the last release).

Do you need this package?

If your MCP client can connect to a remote server, you do not: add https://mcp.getmosaic.dev/mcp as a connector (in claude.ai: Settings → Connectors → Add custom connector). It signs you in by email (OAuth) and needs no key and no install.

This package is for clients that only run local (stdio) servers.

Related MCP server: OrigeneMCP

Quick start

pip install mosaic-mcp          # or: uvx mosaic-mcp
export MOSAIC_API_KEY="mk_live_..."
mosaic-mcp                      # stdio MCP server

With Claude Desktop

{
  "mcpServers": {
    "mosaic": {
      "command": "uvx",
      "args": ["mosaic-mcp"],
      "env": { "MOSAIC_API_KEY": "mk_live_..." }
    }
  }
}

With Claude Code

claude mcp add mosaic --env MOSAIC_API_KEY=mk_live_... -- uvx mosaic-mcp

Tools

Tool

Tier

What it does

mosaic_get_coverage_grid

free

A coverage grid for any human gene symbol: every axis with its state; network axes queued until a dossier fetches them

mosaic_verify_claim

free, rate-limited

One claim about a target checked against Mosaic's own independent count, in at most 10 s, with one of seven verdicts (confirmed, contradicted, cannot_verify_same_source, cannot_verify_floor, cannot_verify_unavailable, cannot_verify_unscoped, out_of_scope)

mosaic_audit_claims

paid ($390 an audit)

Every target-level claim in a document (a press release, a filing section, an AI-generated assessment), audited; async

mosaic_request_dossier

paid ($149 a dossier)

A verified dossier for any human target; async

mosaic_get_result

free

The status and result of a dossier (dossier-N) or audit (audit-N) you requested; only your own

The tier, the price and any refusal come from the hosted service; this package decides none of them.

Configuration

Variable

Required

Description

MOSAIC_API_KEY

Yes

Your Mosaic API key (mk_live_…), sent as a Bearer token. Without it every call says so and sends nothing.

MOSAIC_MCP_URL

No

The hosted MCP endpoint. Default https://mcp.getmosaic.dev/mcp.

A call the hosted service could not answer (unreachable, a timeout, a rejected key) is returned as an error saying so: never as an empty or a zero result.

Changelog

Full history in CHANGELOG.md. 3.0.0: a thin client to the hosted service with the five tools of plan v3 §3.1; the bring-your-own-database mode and its eleven tools are retired (each answers with its replacement).

License

Apache 2.0.

Available Tools

44 tools
mosaic_assess_druggabilityA
Read-onlyIdempotent
Inspect

Assess structural druggability of a target from AlphaFold + fpocket.

Returns the top binding pockets (volume, druggability score), pocket count, and a coarse structural_tier (highly_druggable / druggable / challenging / undruggable) along with a plain-English interpretation. This is the structural answer to "is this target small-molecule tractable?" — orthogonal to literature-derived druggability heuristics.

Inputs: params (required). Returns: result

mosaic_clinical_pipelineA
Read-onlyIdempotent
Inspect

Get clinical trial pipeline for compounds targeting a gene.

Returns compounds in clinical development with indications, trial phases, and status from ClinicalTrials.gov data.

Inputs: params (required). Returns: result

mosaic_compare_drugsA
Read-onlyIdempotent
Inspect

Side-by-side comparison of two compounds.

Returns max_phase, first approval year, molecule type, shared targets, unique targets per side, and per-target potency (pChEMBL) for both. Use this for competitive analyses like "osimertinib vs erlotinib" or "imatinib vs dasatinib".

Inputs: params (required). Returns: result

mosaic_compare_targetsA
Read-onlyIdempotent
Inspect

Side-by-side comparison of 2-5 drug targets.

Returns compound counts, patent counts, paper counts, best IC50, max clinical phase, attractiveness scores, and momentum for each target.

Inputs: params (required). Returns: result

mosaic_competitive_landscapeA
Read-onlyIdempotent
Inspect

Get the full competitive landscape for a drug target.

Multi-hop traversal: Target <- Compounds, Target <- Patents -> Organizations. Shows which pharma/biotech companies are active on this target, how many patents and compounds each has, and overall competitive intensity.

Inputs: params (required). Returns: result

mosaic_compound_analogsA
Read-onlyIdempotent
Inspect

Get structural analogs of a compound with Tanimoto similarity.

Returns analogs with similarity scores, shared scaffolds, and their activity against targets. Useful for SAR analysis and lead optimization.

Inputs: params (required). Returns: result

mosaic_compound_polypharmacologyA
Read-onlyIdempotent
Inspect

Get the polypharmacology profile of a compound — all targets it interacts with.

Shows every target the compound has semantic relations with, the mechanism of action for each (inhibits, agonizes, degrades, etc.), and evidence counts. Useful for understanding off-target effects, repurposing potential, and selectivity from a semantic (not just activity) perspective.

Inputs: params (required). Returns: result

mosaic_compound_selectivityA
Read-onlyIdempotent
Inspect

Get the selectivity profile of a compound across all targets.

Shows activity values against every target the compound has been tested on. Critical for assessing off-target effects and safety liability.

Inputs: params (required). Returns: result

mosaic_drug_repurposingA
Read-onlyIdempotent
Inspect

Find new indications where a compound's primary targets are implicated.

Returns indications where the compound's targets have supporting evidence but the compound itself is not yet clinically active. Ranked by a simple target-support × avg-evidence score. Useful as a starting point for repurposing hypotheses — not a substitute for clinical review.

Inputs: params (required). Returns: result

mosaic_emerging_signalsA
Read-onlyIdempotent
Inspect

Targets whose recent literature/patent activity significantly exceeds their own prior baseline (z-score > 2).

Simple statistics over monthly counts — no ML. Use to spot targets heating up before they crowd. sparkline is recent monthly counts.

Inputs: params (required). Returns: result

mosaic_evidence_mapA
Read-onlyIdempotent
Inspect

Get the full evidence landscape for a drug target from semantic extraction.

Shows all relation types (validation, resistance, biomarker, safety, efficacy, expression, pathway, drug target ID) broken down by source type (paper vs patent), with confidence stats and top evidence snippets per relation type.

Use this to understand the strength and breadth of evidence for a target.

Inputs: params (required). Returns: result

mosaic_find_opportunitiesA
Read-onlyIdempotent
Inspect

Find underexplored high-potential drug targets — white-space opportunities.

Identifies targets with high scientific validation but low competitive intensity. These are the best opportunities for novel drug programs where the biology is strong but Big Pharma hasn't crowded the space.

Ranked by opportunity_score = validation × (1 - competition) × momentum_boost.

Inputs: params (required). Returns: result

mosaic_find_similar_targetsA
Read-onlyIdempotent
Inspect

Structurally similar targets to a given gene, ranked by Foldseek TM-score over the AlphaFold PDB corpus.

Returns the top-k neighbours (default 10) with neighbour metadata (name, target_class, druggability_tier) and the structural-similarity metrics (tm_score normalised over query length, alntmscore over alignment length, evalue, lddt, rmsd). Use for paralog / fold-analog discovery, scaffold-hopping target ideation, and cross-family chemistry repurposing.

Source table is populated by scripts/build_foldseek_index.py. When the index hasn't been built yet, returns an empty result with a populate-hint rather than failing — degrades cleanly on fresh DBs.

Inputs: params (required). Returns: result

mosaic_find_undruggable_targetsA
Read-onlyIdempotent
Inspect

Find validated targets that are structurally hard to hit with small molecules.

Returns targets in the 'challenging' or 'undruggable' tier (or with a top fpocket druggability score below the threshold), plus their pipeline gap signals (compound count, approved drug count, validation count) and a suggested modality (PROTAC / glue, biologic / PPI, fragment-based, or allosteric SBDD). This is the white-space tool for new-modality programs.

Ranked by opportunity_score = validation × (1 - top_pocket_score) × (1 - competitive_intensity).

Inputs: params (required). Returns: result

mosaic_get_target_compoundsA
Read-onlyIdempotent
Inspect

Get compounds active against a specific drug target.

Returns compounds with activity data (IC50, Ki, etc.) sorted by potency.

Inputs: params (required). Returns: result

mosaic_get_target_papersA
Read-onlyIdempotent
Inspect

Get scientific papers mentioning a specific drug target.

Returns publications from PubMed/OpenAlex with titles and dates.

Inputs: params (required). Returns: result

mosaic_get_target_patentsA
Read-onlyIdempotent
Inspect

Get patents mentioning a specific drug target.

Returns patent filings with titles, dates, and assignee organizations.

Inputs: params (required). Returns: result

mosaic_get_target_profileA
Read-onlyIdempotent
Inspect

Get a comprehensive intelligence dossier for a drug target.

Returns UniProt biology, target scores, SAR summary, disease associations, validation evidence, pathways, PPIs, competitive landscape, clinical pipeline, and publication momentum. This is the primary tool for any target question.

Inputs: params (required). Returns: result

mosaic_get_target_structureA
Read-onlyIdempotent
Inspect

Get the AlphaFold structural snapshot for a drug target.

Returns AlphaFold model URLs (PDB / CIF / PAE), per-residue confidence summary (mean pLDDT, fractions of residues at high / confident / low confidence, disordered fraction), and protein length. Useful for SBDD scoping, disorder/IDR risk, and confidence-aware target triage. Pair with mosaic_assess_druggability for binding-pocket scoring.

Inputs: params (required). Returns: result

mosaic_indication_landscapeA
Read-onlyIdempotent
Inspect

Get the full therapeutic landscape for a disease indication.

Shows all targets implicated in this indication, compounds in development, and clinical status.

Inputs: params (required). Returns: result

mosaic_kg_statsA
Read-onlyIdempotent
Inspect

Get overall statistics for the Mosaic knowledge graph.

Returns entity counts (targets, compounds, papers, patents), semantic relation totals and breakdown by type, coverage metrics, and ChEMBL activity counts. Use this to understand the scope and coverage of the KG.

Inputs: no named parameters. Returns: result

mosaic_kol_finderA
Read-onlyIdempotent
Inspect

Rank the top authors (KOLs) publishing on a target or indication.

Scores by paper volume with recency weighting (papers since 2023 count 2x). Returns person name, paper count, most recent publication date, and affiliated organizations. Useful for advisory board assembly, trial PI scouting, and competitive intelligence.

Inputs: params (required). Returns: result

mosaic_list_indicationsA
Read-onlyIdempotent
Inspect

List all therapeutic indications available in the knowledge graph.

Returns every indication with the number of associated targets. Use this to discover which disease areas are loaded.

Inputs: no named parameters. Returns: result

mosaic_list_subindicationsA
Read-onlyIdempotent
Inspect

List fine-grained oncology sub-indications in the knowledge graph.

Sub-indications are histology- or biomarker-defined cancer subtypes (e.g. 'EGFR-mutant NSCLC', 'triple-negative breast cancer') organized under broader parent indications. Optionally pass parent_indication (id, name, or synonym, e.g. 'lung cancer' or 'NSCLC') to list only its children. Each entry includes the number of linked targets.

Inputs: params (required). Returns: result

mosaic_modality_gapsA
Read-onlyIdempotent
Inspect

Which compound modalities are explored vs absent for a target.

Modality is a heuristic SMILES classification (small_molecule, covalent, degrader, macrocycle, peptide_like) over the top-ranked compounds — partial coverage by design; unclassified is reported explicitly. Target-level only: protein_family is not populated, so family-level rollups are unavailable (stated, not silently wrong).

Inputs: params (required). Returns: result

mosaic_org_portfolioA
Read-onlyIdempotent
Inspect

Get a pharma/biotech organization's full portfolio.

Shows which drug targets they're active on, their patent filings, therapy area focus, and competitive positioning. Use to understand what a company is working on and where they're investing.

Inputs: params (required). Returns: result

mosaic_pathway_contextB
Read-onlyIdempotent
Inspect

Get pathway context for a drug target.

Shows which biological pathways the target participates in, other targets in the same pathways, and protein-protein interactions.

Inputs: params (required). Returns: result

mosaic_regulatory_statusA
Read-onlyIdempotent
Inspect

Query openFDA for a drug's approval status, label indications, and adverse events.

Returns: FDA approval dates, brand/generic names, sponsor, product type, route of administration, and a summary count of serious adverse events. Data is live from openFDA — does not require local ingestion.

Inputs: params (required). Returns: result

mosaic_resistance_bypass_mapA
Read-onlyIdempotent
Inspect

Candidate resistance-bypass / escape targets for a given target.

From a deterministic keyword pass over the literature (resistance_relations — GLiREL has no resistance edge type), surfaces targets co-mentioned with the query target in resistance-context abstracts, ranked by a drugability-gap score (strong resistance evidence, low development activity). Hypothesis generator, not evidence — every row carries its source snippet.

Inputs: params (required). Returns: result

mosaic_search_targetsA
Read-onlyIdempotent
Inspect

Search the knowledge graph for drug targets by gene symbol or keyword.

Returns matching targets with basic metadata. Optionally filter by therapeutic indication. Use this as the starting point to explore targets.

Returns: JSON list of matching targets with gene_symbol, name, target_class, and counts of related compounds, patents, and papers.

Inputs: params (required). Returns: result

mosaic_subindication_breakdownA
Read-onlyIdempotent
Inspect

Break a target's oncology associations down by sub-indication.

For the given gene, returns the most relevant cancer sub-indications (e.g. EGFR -> NSCLC subtypes vs. colorectal) with the evidence type and confidence of each link. Complements mosaic_get_target_profile with finer indication granularity.

Inputs: params (required). Returns: result

mosaic_synthetic_lethal_whitespaceA
Read-onlyIdempotent
Inspect

Find synthetic-lethal whitespace: targets functionally coupled to a developed (drugged) target but themselves undeveloped.

For an anchor target with chemical matter, surfaces partners that share STRING protein-protein interactions and/or Reactome pathways with it yet have < 5 patents and no clinical compound — i.e. strong biological coupling, low competitive activity. Ranked by a whitespace score and returned with a deterministic suggested experimental approach.

Scope: co_functionality_proxy is a PPI + shared-pathway proxy for co-essentiality, NOT DepMap co-essentiality (DepMap ingestion is a later sprint). Treat it as a hypothesis generator, not evidence.

Inputs: params (required). Returns: result

mosaic_talent_migrationA
Read-onlyIdempotent
Inspect

Who works on a target, recency-weighted, and what else they work on — a talent-flow signal.

Surfaces the most active researchers on a target (publication-based, using the resolved persons table) and, for each, the other targets they've published on over time — i.e. "people who worked on X now also on Y". Patent inventors are not included (not ingested).

Inputs: params (required). Returns: result

mosaic_target_mechanismsA
Read-onlyIdempotent
Inspect

Get the mechanism-of-action profile for a drug target.

Returns how compounds interact with this target — inhibitors (covalent, allosteric, competitive), agonists, antagonists, degraders (PROTAC). Also shows semantic edge types: validation evidence, resistance mechanisms, biomarker roles, safety concerns, and clinical efficacy signals.

Extracted by GLiREL from paper and patent abstracts.

Inputs: params (required). Returns: result

mosaic_target_networkA
Read-onlyIdempotent
Inspect

Get the full knowledge graph network around a drug target.

Returns all connected entities (compounds, diseases, pathways, organizations, interacting proteins) as nodes and edges. Shows how a target connects to the broader drug discovery landscape.

Useful for understanding the full context of a target and finding non-obvious connections.

Inputs: params (required). Returns: result

mosaic_target_scoresA
Read-onlyIdempotent
Inspect

Get computed attractiveness scores for a drug target.

Returns overall target attractiveness, scientific validation, druggability, competitive intensity, and research momentum (0-1 scale) with direction.

Inputs: params (required). Returns: result

mosaic_target_validationA
Read-onlyIdempotent
Inspect

Get experimental validation evidence for a drug target.

Returns genetic (CRISPR/siRNA), in vivo (animal models), clinical (patient data), and pharmacological validation evidence from literature. Includes specific papers with model systems and outcomes.

Inputs: params (required). Returns: result

mosaic_target_wishlist_addA
Idempotent
Inspect

Request a target Mosaic does not yet cover.

Use this when a gene is outside the current covered set so the operator can prioritise it in the next ingestion batch. Idempotent: re-requesting the same gene/email bumps a counter. Also returns the closest covered targets so the user still gets a useful answer.

Inputs: params (required). Returns: result

mosaic_trial_resultsA
Read-onlyIdempotent
Inspect

Return real ClinicalTrials.gov records (NCT ID, title, phase, sponsor, status).

Filters by any combination of gene symbol, compound name, or indication. Unlike mosaic_clinical_pipeline which can synthesize from max_phase, every row returned here has a real NCT ID and brief title. Use this when you need to cite specific trials or highlight read-outs.

Inputs: params (required). Returns: result

mosaic_watchlist_add_itemA
Idempotent
Inspect

Add a watched entity (target/indication/organization/compound/ relation_type) to a watchlist. Idempotent — re-adding is a no-op.

Inputs: params (required). Returns: result

mosaic_watchlist_createAInspect

Create a watchlist to track targets, indications, orgs, or compounds.

owner_key is the user id, email, or an 'anon:' for anonymous sessions. Returns the new watchlist id to use with mosaic_watchlist_add_item.

Inputs: params (required). Returns: result

mosaic_watchlist_getA
Read-onlyIdempotent
Inspect

Get a watchlist with its items and recent detected events.

Inputs: params (required). Returns: result

mosaic_watchlist_listA
Read-onlyIdempotent
Inspect

List an owner's watchlists with item and recent-event counts.

Inputs: params (required). Returns: result

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A4/5.0

    Scored across 44 tools

    Disambiguation5/5

    Each tool has a clearly distinct purpose, targeting specific aspects of drug discovery like target assessment, compound analysis, competitive landscape, and clinical trials. The descriptions are detailed enough to avoid confusion, even for similar-sounding tools.

    Naming Consistency5/5

    All tools follow the 'mosaic_<verb>_<noun>' or 'mosaic_<noun>_<verb>' pattern in snake_case, with verbs like 'get', 'compare', 'find', 'list', and 'assess'. The naming is predictable and uniform, aiding tool selection.

    Tool Count4/5

    With 44 tools, the server covers a broad domain thoroughly. While the count is high, each tool serves a specific function without redundancy, and the scope justifies the number. A slight reduction could improve manageability, but it's still well-scoped.

    Completeness5/5

    The tool set comprehensively covers drug discovery workflows: target discovery, druggability assessment, compound profiling, competitive intelligence, clinical trials, and more. Gaps are acknowledged and minimal, making the surface complete for its stated purpose.

    Maintenance

    ActivityActive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers