Agents

Turn ideas into experiments.

Delegate execution so your team focuses on the science that matters.

Request a demo →
Vecura — Agent Dashboard

Agent

9 active

Today

KRAS G12C covalent fragments

Docking 1,200 warheads·Analyzing · oncology-2026

PD-1 / PD-L1 cyclic peptides

Planning next steps·Planning · immuno-discovery

GSK-3β CNS leads

ADMET filtering 47 hits·3 models · cns-pipeline

FLT3-ITD repurposing sweep

12 candidates ranked·Complete · aml-discovery

This Week

EGFR T790M selectivity sweep

47 analogs · +12 / −3 · oncology-2026

JAK2 V617F off-target panel

Kinome panel · +89 / −4 · jak-program

Antibody developability triage

32 mAbs · +18 / −2 · biologics

Solubility rescue for LR-104

12 analogs · +6 / −1 · lead-opt

Allosteric pocket search · BRAF

Cryptic site scan · +1 site · braf-allo

⌘↵

Everywhere your science lives

One agent across every surface.

Web App

Plan, run, review — in one workspace.

screen.py

from vecura import Agent

a = Agent()

a.run(

"screen 500 ZINC"

)

Python SDK

Run agents from any notebook or script.

In [3]:vecura · py3.11

vecura.run("dock")

✓ 47 hits · ΔG < −7.5

Jupyter

Drop results straight into your analysis.

#discovery-runs

Vecura

KRAS run done — 12 leads, 3 flagged.

Slack & Teams

Get results where your team already talks.

Understands your science, no matter the target

Vecura learns the structure, the literature, and your team's conventions before it picks up a single tool.

4 subagents · parallel

Fold target — KRAS G12C

Editing files · OpenFold

Dock fragment library

Running · AutoDock-GPU

ADMET sweep on hits

Writing report · ADMETlab 3

Literature cross-check

Pending · PubMed-RAG

Parallel subagents

Independent agents explore your target in parallel — each using the best tool for the question it's answering.

Context · KRAS G12C

“What's known about covalent KRAS G12C inhibitors?”

Searchedinternal · KRAS G12C SAR notes
ReadPDB · 6OIM (Switch-II pocket)
ReadLit · Patricelli 2016 ACS Med Chem
ReadLit · Canon 2019 Nature (AMG 510)
SearchedZINC22 · covalent fragments (R-NH-acrylamide)

Knowledge indexing

Vecura indexes literature, internal SAR notes, and prior runs — citing every source it used to answer.

rules.md

Always check hERG, Ames, CYP3A4 before ranking

Discovery team

Prefer Enamine REAL for fragment libraries

Med chem

Exclude PAINS, Brenk, and assay-interference scaffolds

Lab-wide

Flag any hit overlapping Pfizer 2024 IP space

Legal · all programs

Team rules

Teach Vecura your lab's conventions — preferred libraries, mandatory safety checks, IP exclusions.

Spans the full discovery lifecycle

From clarifying the question, to generating candidates, to chasing down why something didn't work.

Plan · clarifying questions
1

What ΔG cutoff should I use for hit selection?

2

Should I include covalent warheads, or non-covalent only?

3

Run ADMET on all hits, or only those passing ΔG cutoff?

Plan

For ambiguous goals, Vecura asks clarifying questions, agrees on a protocol, then runs it.

Design · 4 analogs generated
VC-A-1042−F → −Cl−9.8
VC-A-1058Pyridine → Pyrimidine−9.5
VC-A-1071Add −OMe at R3−9.2
VC-A-1083Tert-butyl → cyclopropyl−8.9

Design

Propose analogs, optimize for selectivity or ADMET, and explore new scaffolds — all with rationale.

Validate · why did hit fail?

Re-checked docking pose convergence

Top pose stable across 3 seeds

Re-ran ADMETlab with extended endpoints

hERG re-flagged at 5 μM threshold

Searched literature for hERG-prone scaffolds

Quinoline core known liability

Root cause: quinoline scaffold drives hERG risk

Validate

When a hit fails, Vecura traces it back — re-running tools, checking the literature, naming the cause.

Equipped to do real science

Vecura runs tools, ingests your data, and keeps every step reproducible.

vecura · run · isolated

$ vecura.dock(target="6OIM", lib="zinc22")

# Spawning AutoDock-GPU container...

# Docking 1,200 fragments on 4×A100

✓ 47 hits passed ΔG < −7.5 (4m 18s)

$ vecura.admet(hits)

✓ 31 / 47 cleared safety panel

Run tools

Docking, folding, ADMET, MD — Vecura runs them in isolated, GPU-backed containers. No setup.

Compose · @-mention any artifact
Optimize @PDB:6OIM binders. Use @enamine-real and apply our @CNS-rules.md
Compare with @AMG-510.smi reference.

Add context

Point Vecura at exact PDBs, SMILES, sequences, papers, or internal docs with @-mentions.

Run history · KRAS-G12C-2026

Started KRAS G12C run

Jan 8

Added covalent warhead filter

Jan 12

Expanded library to 1,200 fragments

Jan 18

ADMET sweep complete

Yesterday

Filtered hERG-prone scaffolds

3h ago

Top-12 leads exported

Now

Snapshots & reproducibility

Every run is versioned — see exactly what tool, model, and parameters produced any result. Roll back anytime.

Extend with tools and workflows

Give Vecura your existing context, and add custom capabilities for your lab.

Plugin marketplace
Benchling LIMS Sync
Data
Schrödinger Suite Bridge
Modelling
Enamine REAL Catalogue
Compounds
Dotmatics Integration
ELN
GraphPad Prism Export
Analysis

Plugins

One-click connectors to Benchling, Schrödinger, Dotmatics, and the rest of your lab's stack.

Skills · slash commands

/fragment-growth

Grow a fragment with R-group enumeration

/admet-triage

Apply our lab's 6-endpoint safety panel

/selectivity-sweep

Run target across the full kinome panel

/ip-check

Cross-check leads against patent landscape

/synth-tractability

Score retrosynthesis difficulty

Skills

Bundle your lab's SOPs into reusable slash commands. Anyone on the team can invoke them.

MCP · connected sources
Internal SAR DB
connected
PubMed
connected
ChEMBL
connected
Lab Slack
connected
S3 · raw assays
connected

MCP

Connect proprietary tools, internal databases, and custom models via the Model Context Protocol.

From the people running the experiments

Our team spans medicinal chemistry, biology, and comp chem. Vecura is the first tool that actually works for all three without specialised setup. We've gone from hand-rolling pipelines to running entire campaigns through one workspace.

Head of Discovery

Series A biotech

I used to spend half my day formatting outputs between tools. Vecura just handles it — I describe the experiment and get back ranked hits with safety flags already attached.

Computational Chemist

Oncology biotech

The ADMET integration alone saves us from wasted screening cycles. We know which leads are viable before we ever touch the bench.

Drug Discovery Scientist

Mid-size pharma

What used to take a week of script-wrangling — fetching structures, running docking, filtering, summarizing — Vecura does in an afternoon.

Structural Biologist

Academic lab

The agent doesn't just run models — it reasons about which model to use and why. That's the part that surprised me most.

Bioinformatics Lead

Research institute

I can ask it to screen a target, generate a brief, and flag literature conflicts — all in one conversation. It's like having a computational collaborator available at 2am.

Principal Scientist

Drug repurposing startup

Changelog

1.4May 20, 2026Parallel subagents for kinome-scale selectivity sweeps
May 19, 2026ADMETlab 3 endpoints expanded to 220
May 18, 2026Benchling LIMS sync (beta)
1.3May 13, 2026Snapshot rollback across multi-agent runs

Start your next discovery with Vecura

Request a demo