The Boltz API is a molecular and protein design API released by Boltz PBC and the Boltz Team on June 16, 2026. It enables users to invoke the BoltzMol-1 and BoltzProt-1 models within research automation environments. BoltzMol-1 is used for small molecule generation, virtual screening, and initial hit discovery, while BoltzProt-1 handles tasks involving the design of proteins and complexes to bind to specific targets. Researchers can integrate these capabilities into existing analysis code via Python or TypeScript SDKs or CLI tools. Additionally, coding agents such as Codex, Claude Code, and Gemini CLI can delegate molecular and protein design tasks to the API through tool calls. Similar to how GPT converts natural language requests into executable text actions, the Boltz API serves as an interface that translates researchers’ design objectives into computable molecular and protein exploration tasks. Existing computational drug discovery workflows often require connecting different models and scripts at each stage: structure preparation, candidate generation, scoring, and result consolidation. Even when models are open-source, utilizing them requires configuring GPU environments, managing weights, resolving dependency conflicts, and operating servers for batch execution, making immediate adoption difficult for small research teams or automation agents. The key differentiator of the Boltz API is that it wraps molecular generation and protein design models behind a remote interface, reducing operational burdens and enabling not only humans but also scientific agents to execute iterative exploration loops directly. Notably, providing models validated through experimental verification distinguishes this service from simple structure prediction services. However, specific validation data, success rates, benchmark conditions, and per-model input limitations must be verified in the official documentation. In research applications, the API can first automate small molecule hit discovery for disease targets. For example, a researcher prepares the target and design constraints, submits a BoltzMol-1 task via the Python SDK, and then loads the returned candidates into RDKit to calculate molecular weight, cLogP, synthetic accessibility, and structural redundancy. After filtering the primary candidate pool, re-evaluating them with docking or molecular dynamics tools allows the conversion of API-generated results into an experimental priority list. The number of candidates, sampling settings, acceptable constraints, and output formats must be confirmed after reviewing the API documentation. For protein or complex design, the results from BoltzProt-1 can be linked to structure quality assessment and experimental design stages. Researchers input the binding target and design objectives, generate multiple candidates, and then select expression test subjects by performing sequence deduplication, structural confidence review, and filtering for immunogenicity and aggregation propensity. Furthermore, assigning a coding agent the sequence of “candidate generation → result parsing → property calculation → top candidate report generation” allows repetitive design tasks to be organized into reproducible pipelines. However, before applying the API to sensitive projects, conditions regarding the storage, reuse, confidentiality, and intellectual property rights of biological data transmitted via the API must be reviewed in the Supplemental Terms.

The Boltz API is a molecular and protein design API released by Boltz PBC and the Boltz Team on June 16, 2026, enabling users to invoke the BoltzMol-1 and BoltzProt-1 models within research automation environments. BoltzMol-1 is used for small molecule generation, virtual screening, and initial hit discovery, while BoltzProt-1 handles tasks involving the design of proteins and conjugates to bind to a target. Researchers can access these capabilities via Python or TypeScript SDKs or CLI tools with their existing

🔬 Automated Small Molecule Hit Discovery: Generate candidates using BoltzMol-1, then calculate molecular weight, cLogP, and structural redundancy in RDKit. Select top candidates based on predefined physicochemical properties and pass them to docking and cell-based assay pipelines. The number of candidates and generation parameters require verification via the official API schema.