Science Context Protocol — Building a Global Collaboration Network for Autonomous Scientific Agents
SCP is dedicated to breaking down tool barriers in scientific research, enabling researchers to focus on the science itself, and empowering AI agents to become true "experiment designers."
SCP (Science Context Protocol) is an open-source standard protocol designed to accelerate scientific discovery by building a global collaboration network for autonomous scientific agents, connecting heterogeneous scientific resources (software tools, AI models, datasets, workflow engines, lab instruments, etc.).
- 🎯 Vision
- ✨ Core Capabilities
- 🏗️ Architecture
- 🚀 Quick Start
- 🛠️ Tool Ecosystem
- 🎓 Scientific Skills
- 📊 Use Cases
- 🔬 SCP vs MCP
- 🌐 Resources
- 📄 License
- 🙏 Acknowledgments
| Capability | Description |
|---|---|
| 🔗 Protocol-Level Connectivity | Unified description and invocation of 2,200+ scientific resources (tools, models, instruments, etc.) |
| 🧠 Intelligent Orchestration | SCP Hub supports automated planning, execution, and monitoring of multi-step workflows |
| 🧪 Dry-Wet Lab Integration | Seamless integration of computational tools and lab equipment for closed-loop validation |
| 🤖 Multi-Agent Collaboration | Supports multiple AI agents working collaboratively within a unified context |
| 📋 Full Lifecycle Management | End-to-end traceability from registration, planning, and execution to archiving |
| 🔐 Security & Access Control | Fine-grained authentication and authorization mechanisms based on experiments |
SCP adopts a Hub-Spoke Architecture:
┌─────────────────────────────────────────────────────────────┐
│ SCP Client │
│ (Researchers / AI Scientists) │
└───────────────────────┬─────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────┐
│ SCP Hub (Central Orchestrator) │
│ Intent Parsing · Workflow Generation · Task Scheduling │
│ · Permission Management │
└───────────────────────┬─────────────────────────────────────┘
│
┌───────────────┼───────────────┐
▼ ▼ ▼
┌──────────────┐ ┌──────────────┐ ┌──────────────┐
│ SCP Server │ │ SCP Server │ │ SCP Server │
│ (Edge Node) │ │ (Edge Node) │ │ (Edge Node) │
└──────┬───────┘ └──────┬───────┘ └──────┬───────┘
│ │ │
Tools/Models Databases/ Compute
Instruments Resources
- SCP Hub: Central orchestrator responsible for task scheduling and permission management
- SCP Server: Edge node interfacing with local resources
- SCP Client: User interface for researchers or AI scientists
# Clone the repository
git clone https://github.com/InternScience/scp.git
cd scp
# Install dependencies
pip install -e .Set up your own SCP Server and Hub for full control over deployment and management.
Visit SCP Square — we provide a managed SCP Hub where you can submit your own SCP Server to the square for cross-platform discovery and access.
SCP has integrated 2,200+ scientific tools spanning multiple disciplines:
| Domain | Share | Representative Tool Types |
|---|---|---|
| 🧬 Biology & Related Technologies | 45.9% | Genome analysis, protein structure prediction |
| ⚛️ Physics | 21.1% | Quantum computing, materials simulation |
| 🧪 Chemistry | 11.6% | Molecular docking, reaction prediction |
| 🔧 Mechanics & Materials Science | 8.7% | Finite element analysis, molecular dynamics |
| 📐 Mathematics | 8.0% | Symbolic computation, optimization algorithms |
| 💻 Information Science & Computing | 4.6% | Machine learning, data mining |
Building upon tool standardization, SCP further introduces the concept of "Scientific Skills (SCP Skills)" — the first batch of 206 scientific skills is now available in the
skills/directory.
Each Skill is an independently callable, freely composable "research instruction" that orchestrates multiple databases, computational models, and even device drivers behind the scenes, forming a reusable, standardized experiment template.
For Researchers: No need to deal with complex tool manuals and code debugging — focus your energy on the scientific questions themselves For AI Agents: Skills become freely stackable "LEGO bricks," enabling agents to evolve from "tool users" to "experiment designers"
All 206 skills are organized into 8 major scientific domains:
| Domain | Skills | Representative Examples |
|---|---|---|
| 💊 Drug Discovery & Pharmacology | 71 | Target identification, ADMET prediction, virtual screening, molecular docking, drug safety & repurposing, clinical pharmacology |
| 🧬 Genomics & Genetic Analysis | 41 | Variant pathogenicity, cancer genomics, population genetics, rare disease, virus genomics, epigenomics |
| 🧬 Protein Science & Engineering | 38 | Structure prediction (ESMFold/AlphaFold), binding sites, mutation analysis, antibody & peptide design, enzyme engineering |
| 🧪 Chemistry & Molecular Science | 24 | Structure analysis, molecular fingerprints, SAR, material composition, natural products, metabolomics |
| ⚙️ Physics & Engineering Computing | 18 | Circuit analysis, thermodynamics, optics, electromagnetics, crystallography, unit conversion |
| 🔬 Lab Automation & Literature Mining | 7 | Protocol generation, PDF extraction, PubMed search, scientific literature mining, meta-analysis |
| 🌍 Earth & Environmental Science | 5 | Atmospheric science, wind energy assessment, seawater properties, oceanographic calculations |
| 📊 Biomedical Database Integration | 2 | OpenTargets disease-target queries, experimental data processing |
💊 Drug Discovery & Pharmacology — 71 Skills (click to expand)
admet_druglikeness_report · affinity_maturation · alanine_scanning_pipeline · antibody_drug_development · atc_drug_classification · boltz2-binding-affinity · cancer_therapy_design · chemical_safety_assessment · clinical_pharmacology_report · clinical_trial_drug_profile · combinatorial_chemistry · comparative_drug_analysis · compound_to_drug_pipeline · disease-reversal-prediction · disease_compound_pipeline · disease_drug_landscape · disease_knowledge_graph · drug-screening-docking · drug_indication_mapping · drug_interaction_checker · drug_metabolism_study · drug_repurposing_screen · drug_safety_profile · drug_target_identification · drug_target_structure · drug_warning_report · drugsda-admet · drugsda-compound-retrieve · drugsda-data-valid · drugsda-denovo-sampling · drugsda-dleps · drugsda-drug-likeness · drugsda-esmfold · drugsda-file-transfer · drugsda-linker-sampling · drugsda-mol-properties · drugsda-mol-similarity · drugsda-mol2mol-sampling · drugsda-p2rank · drugsda-peptide-sampling · drugsda-prosst · drugsda-rgroup-sampling · drugsda-target-retrieve · enzyme_inhibitor_design · epigenetics_drug · fda-drug-risk-assessment · gene_to_drug_pipeline · gene_variant_drug_nexus · infectious_disease_analysis · lead_compound_optimization · molecular_docking_pipeline · one_health_analysis · opentargets-disease-target · orphan_drug_analysis · pandemic_preparedness · pediatric_drug_safety · personalized_medicine · pharmacogenomics_analysis · pharmacokinetics_profile · polypharmacology_analysis · precision_oncology · protein_drug_interaction · smiles_comprehensive_analysis · structural_pharmacogenomics · substance_toxicology · systems_pharmacology · toxicity_assessment · variant-pharmacogenomics · virtual_screening · gene-knowledge-integration · molecular-property-profiling
🧬 Genomics & Genetic Analysis — 41 Skills (click to expand)
biomarker_discovery · biosample_genomics · chromosome_analysis · comprehensive-variant-annotation · cross_species_genomics · epigenomic_landscape · gene_disease_association · gene_expression_atlas · gene_family_evolution · genetic_counseling_report · genome_annotation · go_term_analysis · microbiome_genomics · mouse_model_analysis · multispecies_gene_analysis · ncbi-gene-retrieval · ncbi_gene_deep_dive · organism_classification · phenotype-by-hpo-id · population_genetics · rare_disease_genetics · region-gene-elements · regulatory_region_analysis · snp_functional_analysis · tcga-gene-expression · tissue_specific_analysis · transcriptome_analysis · ucsc_genome_exploration · variant-clinical-significance · variant-cross-database-ids · variant-functional-prediction · variant-genomic-location · variant-gwas-associations · variant-population-frequency · variant_pathogenicity · virus_genomics · gene_comprehensive_lookup · gene_therapy_target · ensembl-sequence-retrieval · kegg-gene-search · multiomics_integration
🧬 Protein Science & Engineering — 38 Skills (click to expand)
alphafold_structure_pipeline · antibody_target_analysis · binding_site_characterization · blast_protein_analysis · comprehensive-protein-analysis · disease_protein_profiling · dna-rna-sequence-analysis · enzyme_engineering · full_protein_analysis · interproscan-domain-analysis · interproscan_pipeline · metabolomics_pathway · molecular_visualization_suite · mutation_impact_analysis · peptide-properties-calculation · protein-blast-search · protein-properties-calculation · protein_classification_analysis · protein_complex_analysis · protein_database_crossref · protein_engineering · protein_function_annotation · protein_interaction_network · protein_property_comparison · protein_quality_assessment · protein_similarity_search · protein_solubility_optimization · protein_structure_analysis · proteome_analysis · string-ppi-enrichment · structural_homology_modeling · synthetic_biology_design · uniprot-protein-retrieval · uniprot_deep_analysis · lab_protocol_from_literature · code_execution_analysis · molecular-descriptors-calculation · web_literature_mining
🧪 Chemistry & Molecular Science — 24 Skills (click to expand)
aliphatic_ring_analysis · bioassay_analysis · cas_compound_lookup · cell_line_assay_analysis · chembl-molecule-search · chemical-mass-percent-calculation · chemical-structure-analysis · chemical_patent_analysis · chemical_property_profiling · chemical_structure_comparison · compound-name-retrieval · compound_database_crossref · functional_group_profiling · molecular-format-conversion · molecular-properties-calculation · molecular-similarity-search · molecular_fingerprint_analysis · natural_product_analysis · polymer_property_analysis · pubchem-smiles-search · pubchem_deep_dive · smiles-to-cas-conversion · substructure_activity_search · material-density-volume-calculation
⚙️ Physics & Engineering Computing — 18 Skills (click to expand)
buoyancy-acceleration-calculation · capacitance-calculation · electrical_circuit_analysis · electromagnetic_analysis · energy_conversion · experimental_data_processing · geometric-volume-calculation · geometry_trigonometry · length_measurement · measurement-error-analysis · mobility_analysis · nuclear_physics · optical-frequency-calculation · optics_analysis · signal_processing · statistical_error_analysis · thermal_analysis · unit_conversion_suite
🔬 Lab Automation & Literature Mining — 7 Skills (click to expand)
biomedical-web-search · meta-analysis-execution · protocol-extraction-from-pdf · protocol-generation · protocol-to-executable-json · pubmed-article-search · scientific-literature-search
🌍 Earth & Environmental Science — 5 Skills (click to expand)
atmospheric-science-calculations · oceanographic-seawater-properties · seawater-freezing-temperature · seawater-sound-speed-calculation · wind-site-assessment
📊 Other — 2 Skills
seismic-waveform-processing · unit-conversion-nanoscale
- 📦 206 Skills in total, each containing a complete
SKILL.mdwith documentation, test cases, and runnable Python code - ✅ All skills have been tested against live SCP endpoints and verified to work
- 🔧 250+ distinct tools orchestrated across 31 SCP Servers
- 🌍 Covering 8 major scientific domains and 112+ sub-disciplines
- 🔑 Each skill includes authentication setup — just replace
<YOUR_SCP_HUB_API_KEY>with your own key applied and generated in each SCP server page listed in SCP Square
💡 Want more? If you'd like any useful models or tools to be available as 24/7 accessible APIs, just let us know by submitting an issue, and we'll deploy them for you.
Please refer to the SCP Skills Tutorial.
| Scenario | Description |
|---|---|
| 📝 Automated Experiment Protocol Design | Generate executable experiment protocols from natural language objectives |
| 📄 PDF Protocol Auto-Reproduction | Extract experiment steps from PDFs and execute them automatically |
| 💊 AI-Driven Molecular Screening | Integrated QED scoring, ADMET prediction, and molecular docking |
| 🧪 Dry-Wet Integrated Protein Engineering | Closed-loop workflow from sequence design to experimental validation |
📖 For detailed case studies, please refer to our Technical Report or User Cases
| Feature | MCP | SCP |
|---|---|---|
| Protocol Standardization | General tool invocation | Structured, complete scientific experiment workflows |
| High-Throughput Experiment Support | No built-in experiment management | Supports batch experiments and context management |
| Multi-Agent Collaboration | Point-to-point communication | Centralized orchestration and task distribution |
| Wet-Lab Equipment Integration | Requires custom adapters | Standardized device drivers and interfaces |
| Resource | Link |
|---|---|
| 🛠️ SCP Tool Square | Explore 2,200+ integrated tools |
| 🚀 SCP Skills Tutorial | Docker setup, model switching & skills usage guide |
| 📖 Documentation (Chinese) | Detailed user guide |
| 📄 Technical Report | SCP design and experiment details |
| 💬 Community | Discussions and Q&A |
This project is open-sourced under the Apache License 2.0.
SCP is developed by Shanghai AI Laboratory. We thank the open-source community for their support.
If you use SCP in your research, please cite our technical report:
@article{jiang2025scp,
title={SCP: Accelerating Discovery with a Global Web of Autonomous Scientific Agents},
author={Jiang, Yankai and Lou, Wenjie and Wang, Lilong and Tang, Zhenyu and Feng, Shiyang and Lu, Jiaxuan and Sun, Haoran and Pan, Yaning and Gu, Shuang and Su, Haoyang and others},
journal={arXiv preprint arXiv:2512.24189},
year={2025}
}Made with ❤️ by Shanghai AI Laboratory
