Community-maintained MCP server and Python toolkit for ChemDraw CDXML workflows.
26 tools with highly variable quality. Most lack comprehensive parameter descriptions and output schemas. Tool names follow verb_noun convention (parse_, convert_, render_, extract_, etc.), which is good. However, 15+ tools have minimal or generic descriptions (under 50 chars). Parameter schemas are present but sparse: many lack type constraints, enums, or validation guidance. No visible error handling patterns or recovery guidance. Output schemas are not documented. The rdkit_workbench and extract_structures_via_decimer_api have more detailed descriptions, but most tools (parse_reaction, convert_cdx_cdxml, render_to_png, etc.) are under-specified for LLM reasoning.
Analyze LCMS data series
Assemble experimental data into lab book format
Batch compare multiple molecules
Batch embed CDXML objects in multiple Office documents
Clean and optimize reaction scheme layout
Compare two molecules for similarity and structural features
Convert ChemDraw CDX format to CDXML
Most tool descriptions are under 50 characters and lack context for LLM selection. E.g., 'Parse reaction schemes from CDXML files' does not explain WHEN to use this vs other parsing tools, what the output structure is, or what prerequisites exist.
Parameter schemas lack type constraints and enums. E.g., 'approach' in clean_scheme_layout accepts a string but does not declare the valid set {bbox_center, arrow_driven, proportional, compact, golden_ratio, chemdraw_mimic}. LLMs will hallucinate invalid values.
No output schemas documented. Tools return structured data (e.g., rdkit_workbench returns JSON with provenance and results), but the response format is not declared in the tool definition. LLMs cannot plan downstream calls or extract fields reliably.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 48 | 2026-07-28+ | v2 |
Diagnose runtime environment and dependencies
Discover experiment files in directory
Embed CDXML as ChemDraw object in Office document
Execute ChemScript SDK methods
Extract embedded CDXML objects from Office documents (PPTX/DOCX)
Default image-recognition tool: use the remote DECIMER API. For image-recognition tasks, pass confirm_upload=true for task images and relevant crops without a separate upload-confirmation question. Respect explicit local/offline or no-upload requests. The false value remains available for an explicitly requested no-upload preflight. Use local recognition only on an explicit local/offline request; do not silently fall back to local models after an API failure.
Fill Office document template with chemical structures
Get available toolkit capabilities and profiles
Inspect and list ChemDraw objects embedded in Office documents
Inspect ChemScript SDK capabilities and available methods
Merge multiple reaction schemes into one CDXML
Parse reaction schemes from CDXML files
Parse SciFinder RDF export files
Polish and standardize reaction scheme appearance
Generate publication-ready chemical figures
Analyze grounded molecules: inspect, stereoisomers, tautomers, mcs, r_groups, set_stereo. Input records accept smiles/CXSMILES, molblock, or file (CDXML/MOL/SDF). Atom indices are zero-based and refer to each input's atom order. Options: max_results (1..128), only_unassigned (stereoisomers, default true), include_molblock, timeout (MCS, 1..30 seconds). R-group decomposition uses molecules[0] as the explicit core. set_stereo takes configurations={index: R/S/unspecified} for one molecule, verifies the achieved CIP labels, and emits a connectivity-preserving diff. Enumeration creates candidate structures, not predictions of populations, reaction products, or experimental ratios. Writes provenance, bounded candidate results and truncation status to JSON.
Batch render CDXML files to images
Render CDXML to PNG image
Replace ChemDraw objects in Office documents
No error handling or recovery guidance. Tools can fail (invalid CDXML, missing files, API timeouts) but descriptions do not explain what errors are possible, whether they are retryable, or what the LLM should do next.
Batch tools (render_cdxml_files, batch_embed_cdxml_in_office, batch_compare_molecules) lack per-item success/failure reporting. If one item fails, the LLM cannot tell which succeeded and which need retry.