Skip to main content
Glama
Mannich MCP Server

Mannich MCP Server

by chemagents

mannich-mcp-server

An MCP server for reproducing and auditing the analyses in:

Vasechkin, K. V.; Nikitin, I. D.; Fedorov, M. V. A curated comprehensive dataset of asymmetric organocatalytic Mannich reaction. Scientific Data 2026. https://doi.org/10.1038/s41597-026-08287-9

The server recomputes the data descriptor from the authors’ manually curated table: 4,015 reactions, 27 source fields and 153 normalized publication DOIs. It preserves reaction records, conditions and axial stereochemical labels. It does not extract reactions from papers again. The dataset and scientific implementation are bundled in the pinned dependency; the analyses run on CPU.

Scientific code and source data are shared through chem-repro-suite, pinned to version 0.4.4, commit 4be1e94c912db7a2790c831fb01d098ad290a88a. This repository supplies the MCP entry point, deployment configuration and checks. Tools operate on the fixed article dataset; they do not accept arbitrary uploaded datasets.

Tools

Tool

What it returns

mannich_dataset_overview

Record counts, molecular identities, missingness and duplicate-input audit

mannich_enantioselectivity_distribution

ee-to-absolute-ΔΔG‡ conversion, distributions and skewness

mannich_reaction_conditions

Temperature, solvent, time, yield and catalyst-loading statistics

mannich_component_classification

Ordered SMARTS classes, source-label audit and representative structures

mannich_chemical_space

Morgan fingerprint t-SNE maps with source identities and class labels

mannich_scaffold_diversity

Murcko scaffold populations, frequencies and acyclic exclusions

mannich_selectivity_heatmaps

Median ee and observation counts by catalyst–substrate class

mannich_catalyst_structure_audit

Deposited SDF geometry and point/axial stereochemistry checks

mannich_regenerate_catalyst_structures

Separate source-protocol and geometry-verified conformer generation

mannich_reaction_space

Explicit reconstruction of the supplementary reaction PCA

Exact parameters and output schemas are in api/tools.json. Each result includes metrics, tables, limitations and provenance. Downloadable artifacts have a URL, byte count and SHA-256; full CSV files accompany table previews. success means the computation completed, not that every paper claim matched.

For example, ask CoScientist:

Recompute the composition of the published Mannich dataset, the ee-to-absolute-ΔΔG‡ conversion, reaction-condition statistics and catalyst/substrate classes. Preserve every record and axial label, list missing values and discrepancies separately, and return the tables and figures.

See REPRODUCTION_QUESTIONS.md for further questions and the tools expected to answer them.

Related MCP server: FLASK-tools

Reproduction fidelity

Result

Recomputed result / comparison

Reactions / publications

4,015 / 153 normalized DOIs

Catalysts / electrophiles / nucleophiles

468 / 578 / 538; catalyst identity includes the axial label

Median ee / yield / reaction time

89.4% / 83% / 24 h

ee / absolute ΔΔG‡ skewness

−1.547834 / 0.102246; stored conversion agrees to numerical precision

Catalyst class labels

No disagreements with the deposited labels

Catalyst / electrophile / nucleophile scaffolds

207 / 114 / 112; the supplement reports 204 catalyst scaffolds

The 468 catalyst identities correspond to 462 isomeric SMILES; ordinary Morgan bits do not encode axial R/S labels. Solvent counts depend on the definition: 51 non-mixture labels excluding neat conditions, 77 raw strings or 50 grouped categories. These are not interchangeable molecular-identity counts.

Heatmaps mask cells with fewer than 5 observations in paper mode and fewer than 4 in notebook mode. They describe literature observations, not prospective catalyst performance. The catalyst SDF audit retains unresolved or conflicting axial assignments. Reaction PCA is a declared reconstruction because the supplement does not specify the exact fingerprint and PCA settings.

Run locally

Requires Git, Python 3.12 and uv:

git clone https://github.com/chemagents/mannich-mcp-server.git
cd mannich-mcp-server
uv sync --frozen --python 3.12
uv run mannich-mcp-server

The default port is 7343. This starts a local development instance; no LLM API key is required by the MCP server.

Attach to CoScientist

Deploy once on a host reachable by CoScientist and register its Streamable HTTP /mcp endpoint. The same service exposes result files at /files/…. End users then ask questions through CoScientist.

No shared hosted endpoint is currently published. Set MCP_BASE_URL to the actual deployment address, without /mcp, and generate the configuration:

uv run python scripts/configure_endpoint.py --base-url "$MCP_BASE_URL"

Deployment, Docker and registration instructions cover startup, artifact URLs, optional workers and verification from the client network. server.json records the integration contract and deployment status.

Tests

python -m unittest discover -s tests -v
uv run python scripts/smoke_server.py --study mannich

CI checks configuration, startup, tool discovery and artifact downloads in Python and Docker. These checks do not establish full scientific reproduction or acceptance of every CoScientist answer. The English questions define the system evaluation. Current end-to-end runs use local CoScientist fixes for MCP result capture, source access and reporting; they do not certify the unmodified upstream version. Scientific findings and source limitations are reviewed separately from successful execution.

License / data

Code: MIT. Source provenance and dataset licenses are listed in the data manifest and NOTICE. Please cite the article when using its data or results.

Tool Schema Changelog

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

Loading tool schema changelog…

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      B
      maintenance
      Chemical safety intelligence for AI-assisted experiment design. 18 MCP tools for compatibility checks, hazard analysis, PPE, storage, waste disposal, mixing order, exposure limits, transport classification, regulatory compliance (8 jurisdictions), and signed audit reports. Supports 4,200+ chemicals in 5 languages.
      MIT
    • F
      license
      Not graded
      quality
      A
      maintenance
      Provides a collection of MCP servers for computational chemistry tasks including molecular generation and retrosynthesis. Also offers property prediction and molecule pricing capabilities.
      -
    • A
      license
      Not graded
      quality
      C
      maintenance
      An MCP server that gives LLMs native access to cheminformatics and molecular ML tools, enabling molecular structure manipulation, descriptor calculation, ML model training, and analysis report generation through natural conversation.
      37
      MIT
    • F
      license
      Not graded
      quality
      D
      maintenance
      A self-driving cheminformatics MCP server that dynamically exposes a growing library of RDKit-based molecular analysis tools (fingerprints, descriptors, substructure matching, drug-likeness filters, and more) as MCP tools, with each skill autonomously implemented and tested by an agent loop without human intervention.
      -