Pricing

AI features measure usage in tokens — the unit AI models use to bill text in and out. A typical property prediction uses ~500 tokens; a retrosynthesis suggestion uses 2,000–5,000 tokens. Standard users get 500,000 tokens per day and 10 million per month — enough for several hundred analyses. Your daily budget resets at midnight UTC. You can track your usage in Settings.

Yes. Cancellation is one click in Settings → Manage subscription, with no contact-required step. You keep access until the end of your current billing period.

Your drawings, documents, import/export, and RDKit computations are still available — only AI features pause until your budget resets.

If you have an institutional email (.edu or another verified academic domain), you qualify for 50% off Standard ($9/mo or $90/yr). Click 'Academic researcher? Get 50% off' on the Standard card to start verification.

Yes — Standard subscribers can opt into BYOK in Settings → API Key. Your own key replaces the platform key for AI requests; usage doesn't count against your token budget. Student tier uses the platform key only.

You retain full ownership of every structure, document, and conversation you create. You can export everything any time via Settings → Export all my data (a single ZIP archive). We never train AI models on user-created chemistry. See our Privacy Policy for full details.

Property predictions (MW, logP, HBD, HBA, TPSA, Lipinski) run on RDKit 2024.3.5. IUPAC naming and name-to-structure conversion run via the pyopsin 0.2.4 wrapper around OPSIN. Both are deterministic — given the same SMILES, you'll get the same answer every time. AI suggestions (reaction prediction, retrosynthesis, naming for novel structures) are clearly distinguished from these computed values in the result cards.

Every tool-result card carries a confidence label. Property cards (calculate_properties) are RDKit-computed and labeled 'Computed.' Reaction prediction, retrosynthesis suggestions, and naming for unknown structures are labeled 'AI Suggested' with a confidence indicator — sanity-check these against your domain expertise or experimental verification before use.

CDXML import is supported (Ketcher's built-in parser); export to CDX/CDXML is on the roadmap. For now, exports use SMILES, MOL V2000, PNG, and SVG — all of which open in ChemDraw via Insert → File. Native CDXML export is tracked on our public roadmap.