01
Explore
Search literature, databases, and public repositories
Literature search · Compound databases · Protein & gene databases
500+ scientific tools · one account
fold, dock, screen and plan the synthesis — in one question.
ask Ivy and it plans the run, submits each job and feeds its output into the next — from the browser, the ivybloom terminal, curl, or your agent over MCP — or curl a single job yourself.
web chatplans a run
From sequence to synthesis · Press Enter to start
start with up to $1 in guest compute per day — no account or card; sign in for saved sessions, higher quota and the full catalog.
each step's output feeds the next, and every step is a row in a hash-chained record you can verify.
explore a few of our 500+ tools
.bloom* studioone job
one tool job from a form, no code — signed in, experimental. start a project
install
npm install -g @ivybiosciences/ivybloom-cli && ivybloom auth login
direct submitone job
ivybloom run esmfold protein_sequence=MKFLILLFNILCLFPVLAADNH --wait
scripted / batchone job each
ivybloom batch submit jobs.yaml
TUI chatplans a run
ivybloom
MCP serverplans a run or one job
ivybloom mcp serve
requires a free account — for a terminal, scripts and CI. get one free
direct submitone job
curl -X POST https://www.ivybiosciences.com/api/v1/jobs -H "Authorization: Bearer $IVY_API_KEY" -H "Content-Type: application/json" -d '{"tool_name":"esmfold","tool_input_parameters":{"protein_sequence":"MKFLILLFNILCLFPVLAADNH"}}'needs a free account and an api key, set as IVY_API_KEY. read the jobs api
ask Ivyplans a run
curl -N https://www.ivybiosciences.com/api/v1/orchestrate -H 'Content-Type: application/json' -d '{"prompt":"find recent papers on EGFR L858R resistance","stream":true}'no account needed · up to $1 in guest compute · 18 calls/day per IP · runs allowlisted GPU tools
onboard your agent
read https://www.ivybiosciences.com/llms-full.txt and follow it to run biomedical tools for me
paste it into your agent; it covers curl, ivybloom and MCP.
/SKILL.mdplans a run or one job
https://www.ivybiosciences.com/SKILL.md
the installable skill; the curl and ivybloom lines in the other tabs are the same requests
MCP configplans a run or one job
{"mcpServers":{"ivybloom":{"command":"ivybloom","args":["mcp","serve"]}}}requires the cli, signed in — adds every tool to your coding agent
same request, same .bloom* registry — an agent pastes what a person would.
- tools
- 500+
- databases
- 50+
- ways in
- 4
4 ways in
the same run, whichever way in.
signed-in runs from the web app, the terminal, curl or MCP all write to the same hash-chained record.
web app
start a runIvy in a browser — no account to start.
http
POST /api/v1/orchestratecurl it with a prompt — no key for the guest tier.
terminal
ivybloomthe terminal app — install it and sign in.
mcp
{"mcpServers":{"ivybloom":{"command":"ivybloom","args":["mcp","serve"]}}}any mcp client — needs the cli, signed in.
four promises, one run
500+ biomedical tools, whoever asks — one run you keep.
you pick the model and how you call it; Ivy plans the run and brings the biology.
- evidence
500+ biomedical tools. every result names the tool, parameters and run that produced it.
see what it computes - artifact
you leave with what the run made — a structure, a ranked table, an export.
see a shared run - record
signed-in runs from the web app, the terminal, curl or MCP leave a tamper-evident record you can check, and so does a guest chat, from the browser tab it is open in, while the guest session lasts.
edit a row and the chain breaks at the verifier. it detects modification; it does not prevent it, and it is not access control.
see how the record works - model
anthropic direct or through openrouter, across the model families your plan includes — or bring your own key once you sign in.
see the models
check a prediction against something measured.
you pick the reference.
- run the prediction, then pull affinity from BindingDB, clearance from the drug label or the deposited pose, and compare.
- gaps stay gaps — when a summary leaves rows out, the run says so instead of guessing them.
a chatbot answers from what it read.
this runs the model and shows you the run.
one is a quote. the other is evidence.
a connector widens what a model can read, never what it can do.
the .bloom* domain
500+ biomedical tools across 6 domains.
search, analyze, predict, design, work across omics and clinical data, then turn the run into a report or an export. the catalog is broad; the entry point stays one question.
02
Analyze
Screen compounds, predict properties, assess safety
ADMET profiling · Safety screening · Cheminformatics
03
Predict
Fold proteins, dock molecules, simulate binding
Structure prediction · Molecular docking · Multi-chain folding
04
Design
Generate molecules, engineer proteins, design biologics
Fragment chemistry · Cloning design · Antibody annotation
05
Omics & Clinical
Analyze genomes, transcriptomes, and clinical data
RNA-seq QC · CRISPR guide design · Differential expression
06
Reports & Automation
Generate passports, export data, automate workflows
Basic passports · Standard export · Literature passports
Six domains, one registry. Signed-in runs from the web app, the terminal, curl or MCP leave a tamper-evident record you can check, and so does a guest chat, from the browser tab it is open in, while the guest session lasts.browse all capabilities
who it's for
it doesn't replace how you work.
whatever you already run keeps running. this shortens the gap between a question and the evidence that settles it — whether that's a supplement you take or a compound you're about to order.
“Which iron supplement is actually the best one?”
“What's actually known about a drug I've been prescribed?”
compare iron supplement forms for absorption, then show me the label and reported side effects for the one I take
- the forms compared side by side, not just brands
- the approved label and filed adverse-event reports
- anything that clashes with what you already take
Shows you what the published evidence says — labels, trials, filed reports. It is not medical advice and won't tell you what to take.
where it fits
leave whenever you like. reach for something else when it fits.
- portable
switch between the model families your plan includes. running the runtime on your own hardware comes with the planned source release; enterprise plans cover self-hosting today.
FSL-1.1 permits internal use, research and teaching outright — it only bars reselling it as a competing service — and each release converts to Apache-2.0 two years after it ships.
see the runtime - hosted gpu
folding and docking run on Modal; the inputs, outputs and citations stay attributable to the run.
see infrastructure - elsewhere
reach for a specialist when the question is narrower — literature reasoning, physics-grade lead optimization, a single folding job.
.bloom* is for the loop between them: a question answered with computed, cited, reproducible evidence in one run.
see the honest side-by-side
ready when you are
start with a question. keep the run.
it's not a replacement for your workflow. it's fertilizer and a greenhouse.
no account for chat or curl. sign in when you need saved sessions, higher quota and the full catalog.
or have your agent paste this — no account
curl -N https://www.ivybiosciences.com/api/v1/orchestrate -H 'Content-Type: application/json' -d '{"prompt":"find recent papers on EGFR L858R resistance","stream":true}'onboard your agent
read https://www.ivybiosciences.com/llms-full.txt and follow it to run biomedical tools for me
paste it into your agent; it covers curl, ivybloom and MCP.