Back to Plugins

Boltz

Predict structures, screen molecules and proteins, and design binders with Boltz from Claude Code.

development
By Boltz
4Updated 1 week agoPythonMIT

Installation

/plugin install boltz@claude-plugins-official

How to install

  1. Open Claude Code in your terminal
  2. Run the installation command above
  3. The plugin will be enabled automatically
  4. Use the plugin's features in your Claude Code sessions

Boltz Skills

Run Boltz biomolecular workflows — structure prediction, binding, molecular and protein screening, and de novo design — directly from your AI coding agent. The skills shell out to the boltz-api CLI to create inputs, estimate cost, submit jobs, and download results.

📖 Full guide: Agent integrations for the Boltz API

Supported agents

AgentHow it's delivered
Claude CodePlugin (boltz) installed from a marketplace
CodexPlugin (boltz) installed from a marketplace
Gemini CLI / AntigravityCLI extension
Claude DesktopLocal MCP server (.mcpb bundle)

Every surface runs the same workflows through the boltz-api CLI.

Skills

SkillWhat it does
boltz-cli-setupInstall, update, verify, and authenticate the boltz-api CLI.
boltz-structure-and-bindingPredict the 3D structure of a defined complex; optionally score binding.
boltz-small-molecule-screenRank a SMILES library against a target.
boltz-small-molecule-designGenerate novel small-molecule binders.
boltz-small-molecule-admeEstimate Tier-1 ADME (solubility, permeability, logD) from bare SMILES.
boltz-protein-screenRank proteins / peptides / antibodies against a target.
boltz-protein-designGenerate novel peptide / antibody / nanobody / custom-protein binders.
boltz-check-statusList and inspect jobs; recover results after an interrupted session.

Prerequisites

All surfaces need the boltz-api CLI on your PATH.

macOS and Linux:

curl -fsSL https://install.boltz.bio/boltz-api/install.sh | sh

Windows PowerShell:

irm https://install.boltz.bio/boltz-api/install.ps1 | iex

Verify it:

boltz-api --version

Then authenticate, using either device-code login or an API key:

boltz-api auth login --device-code
# or
export BOLTZ_API_KEY="your-api-key"

Results download to a boltz-experiments/ directory in your working directory (created automatically). Point any command at a different location with --root-dir.

Install

Claude Code

claude plugin marketplace add boltz-bio/boltz-api-skills
claude plugin install boltz@boltz-marketplace --scope user

Codex

codex plugin marketplace add boltz-bio/boltz-api-skills
codex plugin add boltz@boltz-marketplace

Gemini CLI / Antigravity

gemini extensions install https://github.com/boltz-bio/boltz-gemini-cli
# Antigravity:
agy plugin install https://github.com/boltz-bio/boltz-gemini-cli

Claude Desktop

Download the latest boltz-mcpb-<version>.mcpb from Releases and install it via Settings → Extensions → Advanced settings → Install Extension.

Restart your agent after installing.

Usage

Once installed and authenticated, just describe what you want in natural language — the agent picks the right skill, builds the payload, shows you a cost estimate before submitting, and downloads results when the job finishes. For example:

Predict the structure of this protein–ligand complex and score the binding affinity.

Screen these 200 SMILES against my target and rank them.

See the full guide for detailed workflows and examples.

Local development

Contributions are welcome — outside PRs included. Clone the repo and try the plugin without a persistent install:

claude --plugin-dir ./surfaces/claude-code-cli

The repo is organized as shared source plus per-agent surfaces:

  • core/ — shared skill workflows (core/skills/cli/) and API reference docs (core/references/), symlinked into each surface. Edit here for changes that apply to every agent.
  • surfaces/<agent>/ — per-agent packaging and host-specific files (claude-code-cli, codex-cli, gemini-cli, mcpb). Edit here for changes specific to one host.
  • plugins/ — self-contained, installable copies. Generated — do not edit.

After editing core/ or a surface, regenerate the installable copies:

scripts/generate-surfaces.sh

See CONTRIBUTING.md for the full development, testing, and release workflow.

License

MIT — see LICENSE.

View source on GitHub