Skip to content

RXN-Sandbox

RXN-Sandbox runs local transformer-based product prediction, single-step retrosynthesis and route-tree generation, with notebooks, scripts and an MCP server.

Catalog updated ·

Overview

The repository packages prediction workers and interfaces in containers. Its FastMCP implementation exposes product_prediction, retro_prediction and retro_tree_prediction tools. The supplied models are described as trained on Pistachio 2025Q2.

Git LFS and container tooling are required; tree search can be costly. OpenMDW-1.1 covers supplied model materials, while third-party rights require separate checks. Hosted IBM RXN ends on 2026-10-28; this entry covers the self-hosted sandbox.

Key Features

  • Local forward and retrosynthetic prediction
  • Three verified FastMCP prediction tools

Use Cases

  • Evaluate a local reaction-prediction workflow and optionally connect an assistant through MCP.

How to Use

  1. Read the official instructions and select the documented task and environment.
  2. Obtain only the required data or checkpoints and check their specific terms.
  3. Evaluate a small representative example against references before adapting the workflow. These steps have not been run during curation.

Sources and verification

This entry describes capabilities documented by the linked project sources. Catalog edits are not evidence of a successful run, compatibility test or scientific performance evaluation.

Read the editorial and verification policy

Related resources

ASKCOS

Platforms

ASKCOS is a scientific synthesis-planning platform with retrosynthesis, forward prediction and reaction-analysis modules, offered through a hosted interface and modular deployment documentation.

Python

Reaction Prediction · Retrosynthesis

Graph2SMILES is a graph-to-sequence model for single-step retrosynthesis and chemical reaction outcome prediction.

Open sourcePython

Reaction Prediction · Retrosynthesis

LlaSMol

Model

LlaSMol is a family of chemistry large language models fine-tuned on the SMolInstruct instruction dataset (four released models) covering tasks such as name conversion, property prediction and reaction prediction, with the code under the MIT license.

Open sourcePython

Molecular Property Prediction · Cheminformatics · Reaction Prediction · Retrosynthesis

AiZynthFinder is a Python retrosynthetic planning toolkit that uses neural-network-guided search to propose routes from target molecules to precursors in a configured stock.

Open sourcePython

Retrosynthesis

autodE

Open Source

autodE automates reaction energy profile calculation from reactant and product SMILES, including transition-state location, conformer searching and atom mapping, through wrappers for external electronic-structure codes.

Open sourcePython

Reaction Prediction · Computational Chemistry · Quantum Chemistry

LocalRetro is a research implementation for retrosynthesis prediction using local reaction templates, with workflows for template extraction, model training, testing and decoding predicted reactants.

Python

Retrosynthesis

Related guides

Retrosynthesis

Planning a Retrosynthesis Tool Evaluation

Build a bounded evaluation of retrosynthesis planning, single-step prediction and atom mapping, with explicit inputs, scoring rules and expert review—without treating generated routes as validated laboratory procedures.