ivybloom cli

the same tools, without leaving your terminal.

command-line access to the full ivy biosciences platform. run biomedical workflows, batch process molecules, automate pipelines — and expose every tool to claude code, cursor, or any other ai coding agent via the built-in mcp server.

install the cli read the docs

one go binary · macOS, linux and windows, no runtime dependencies · FSL-1.1-Apache-2.0, apache 2.0 after two years

one platform, many clients

the chat is one client, not the only one.

the CLI gives you a full terminal UI, runs as an MCP server so AI coding agents call your tools, and works as a scriptable command for batch pipelines. the web app, curl to /api/v1/orchestrate, the ivybloom terminal app and any MCP client all reach Ivy; a single tool job can also be dispatched directly with curl to /api/v1/jobs, ivybloom run, or a studio project. an agent starts with curl: POST /api/v1/orchestrate with a prompt — no install, no key, no account. move up only when you need to: api key for your own quota and sessions you can resume (a free account); mcp for an agent client that speaks mcp (the ivybloom cli, installed and signed in); cli for a person at a terminal, or long-running local sessions (the same install and sign-in).

tui + script
A full terminal UI for interactive sessions; the same single binary scripts batch pipelines.
mcp server
Exposes every .bloom* tool over the Model Context Protocol so Claude Code, Cursor, and other agents call them directly.
api + pleach
curl any model endpoint over HTTP, or compose tool chains in code with pleach — the agent runtime .bloom* is built on — with checkpoint, interrupt, and resume.

installation

one binary, one command.

install ivybloom with npm, or download a prebuilt binary. a single binary with zero runtime dependencies.

npm

npm install -g @ivybiosciences/ivybloom-cli

binary

curl -fsSL https://www.ivybiosciences.com/install.sh | sh

quick start

authenticate once, then run.

get up and running in under a minute. parameters are key=value pairs, not flags.

authenticate with your ivy biosciences account
```
$ ivybloom auth login --browser
```

predict protein structure from an amino acid sequence

$ ivybloom run esmfold protein_sequence=MKTLLILAVL...

dock a ligand against a protein target

$ ivybloom run diffdock protein_pdb=target.pdb ligand_smiles='CC(=O)O'

predict ADMET properties for a molecule

$ ivybloom run admet_predict smiles='CC(=O)Oc1ccccc1C(=O)O'

submit a batch of jobs from a manifest file
```
$ ivybloom batch submit molecules.json
```

integration

drop it into the automation you already run.

embed ivybloom into your existing workflows. --json gives structured output for parsing, and --wait blocks until a job completes.

github actions · yaml

- name: run admet screening
  run: |
    ivybloom auth login --api-key ${{ secrets.IVY_API_KEY }}
    ivybloom batch submit candidates.json

python

import subprocess, json

result = subprocess.run(
    ["ivybloom", "run", "esmfold", f"protein_sequence={seq}",
     "--wait", "--json"],
    capture_output=True, text=True,
)
job = json.loads(result.stdout)

shell · bash

#!/bin/bash
# dock one ligand, wait for it, keep the JSON
ivybloom run diffdock protein_pdb=target.pdb \
  ligand_smiles='CC(=O)O' --wait --json > pose.json

mcp server (claude code, cursor) · json

// .mcp.json
{
  "mcpServers": {
    "ivybloom": {
      "command": "ivybloom",
      "args": ["mcp", "serve"]
    }
  }
}

ready to try ivybloom?

install it, sign in, run the first job.

install the cli, authenticate with your account, and start running biomedical workflows from your terminal in under a minute. you leave with what the run made — a structure, a ranked table, an export. signed-in runs leave a tamper-evident record you can check.

install the cli read the docs