chemistry
Deterministic MCP for AI agents: drug discovery, ADMET, docking, LNPs, DoE, retrosynthesis, ICH M7
Install
mcp_config.json
{
"mcpServers": {
"com-covasyn-chemistry": {
"url": "https://mcp.covasyn.com/mcp",
"type": "streamable-http"
}
}
}Documentation
CovaSyn MCP Examples
Production-ready integration snippets for connecting CovaSyn — the deterministic, validated chemistry MCP server for pharmaceutical R&D — to the most common Model Context Protocol clients.
What CovaSyn is in one sentence: the first integrated, deterministic cheminformatics platform for AI agents — not a wrapper around one public database, but 130+ tools across 8 suites (cheminformatics, chem-ADME, toxicology, mass spec, NMR, folding, biologics, platform) with ICH / GxP-aligned audit trail and EU data residency.
Benchmark headline
On the ICLR 2026 MolecularIQ benchmark (Klambauer Lab, JKU Linz, 3,540 symbolically verified chemistry tasks), four frontier LLMs scored 76–92 % with CovaSyn MCP attached, versus 14–41 % at baseline. Numbers, methodology, cost-per-question, and failure modes: covasyn.com/en/benchmark.
When to use CovaSyn (vs. building your own)
- Your AI agent needs to answer chemistry questions correctly, not just plausibly.
- You're a pharma, biotech, or CDMO team that needs ICH M7 / ICH Q1 / Tox21 / CYP450 / docking workflows callable from Claude, ChatGPT, Cursor, or any MCP-aware client.
- You want deterministic, version-pinned tool calls that produce the same result tomorrow as today (regulatory audit-ready).
- You need EU data residency (Hetzner Leipzig) or US hosting on request — your data is never used for training.
CovaSyn is not another LLM. It's the deterministic tool layer that fixes the chemistry-hallucination problem of frontier LLMs.
What's in here
| Example | Client | Description |
|---|---|---|
| claude-desktop/ | Claude Desktop | claude_desktop_config.json snippet |
| cursor/ | Cursor IDE | .cursor/mcp.json snippet |
| vscode/ | VS Code with MCP extension | settings.json snippet |
| python-agent/ | Custom Python agent | minimal stdio-MCP client |
| typescript-agent/ | Custom TypeScript agent | minimal MCP TS client |
| curl/ | Direct HTTP | shell snippets for tool discovery + invocation |
What CovaSyn covers
130 functions across 8 tool families:
covabasic— core cheminformatics (canonicalization, fingerprints, druglikeness)covachem— chem ADME, pKa, profile, comparecovatox— toxicology with ICH M7 batch assessment, Tox21, CYP450covams— mass spectrometry (formula, fragments, impurities, RT)covanmr— NMR analysis (1D, 2D, batch, identification)covafold— protein and RNA folding, binding sitescovabio— antibody, peptide, ADC, mRNA, oligo, siRNAcovaplatform— stability (ICH Q1), DoE, solubility, chromatography, UV/IR, regulatory DB, docking
Full tool reference: covasyn.com/en/mcp
Pricing
- Free tier: 100 credits/week
- Pro: €250/month
- Unlimited: €750/month
- Enterprise: custom (SLA, single-tenant, GxP validation pack)
Compliance posture
CovaSyn is GxP-ready (EU Annex 11, 21 CFR Part 11), GAMP 5 Software Category 4, with ICH M7 (R2) mutagenic impurity assessment and ICH Q1A/E stability workflows. Deterministic, version-pinned, audit-trail-complete.
License
The example snippets in this repository are released under the MIT License — use, modify, and redistribute freely.
CovaSyn itself is commercial software. See covasyn.com for terms.
Contributing
PRs welcome. See CONTRIBUTING.md.
Comparison
How CovaSyn compares to other chemistry MCP options:
Sourced from the repository README.
More in AI & Agents
- PonytailMakes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.109,599
- AgentsMulti-harness agentic plugin marketplace for Claude Code, Codex, Cursor, OpenCode, GitHub Copilot, and Google Antigravity39,079
- Frontend SlidesCreate beautiful slides on the web using a coding agent's frontend skills28,060
- Agent Skills Search ServerSearch and discover Agent Skills from the skills.sh registry. Powered by HAPI MCP server.24,658
- Agency Agents Zh🎭 267 个即插即用的 AI 专家角色 — 支持 Hermes Agent/Claude Code/Cursor/Copilot 等 18 种工具,覆盖工程/设计/营销/金融等 20 个部门。含 52 个中国市场原创智能体(小红书/抖音/微信/飞书/钉钉等)。搭配编排器 agency-orchestrator,一句话即可让多位专家按 DAG 自动协作。19,868
- Watermarks RemoverStrip multi-vendor AI provenance marks: Unicode text hygiene, statistical rewrite hooks, and C2PA/metadata from PNG/JPEG/SVG/PDF/DOCX/HTML/MD17,822