Bioinformatics Cloud
Built for AI Agents
Create a project, start a cloud server, then run your agent on it.
Now working with early users.
How you use it
1
Create a project and start a cloud server
Use a local agent on your laptop or desktop, or the web console.
Paths: Local agent · Web console
2
Connect a remote agent over SSH to that cloud server
Register an SSH key once. For each project, run its setup script (or copy its Host alias config), then connect with Cursor or a CLI.
One Connection. Your Agents Handle the Rest.
Connect a local agent once, or use the web console. Either way you get a project and cloud server.
Read the getting started guide for MCP setup, console steps, and SSH connection.
Authorize once. Then you can:
And your remote agent, running on the cloud server, can:
Why Existing Infrastructure Isn't Enough
AWS is powerful. But...
Your agent shouldn't need to become an AWS administrator. Managing EC2, IAM, S3, VPCs, and credentials takes hours before analysis even starts.
Generic compute platforms are flexible. But...
Bioinformatics needs more than CPUs—large datasets, reference genomes, scientific software, and workflows that understand biological data.
Traditional bioinformatics platforms are human-first.
Agents need APIs, machine-readable state, asynchronous execution, and structured permissions.
Pepkio Cloud is designed around the agent—not the dashboard.
MCP-native, API-first, and built for autonomous workflows with structured safety boundaries.
Built for AI Agents from the Ground Up
Native MCP
MCP is not an add-on. It is the primary interface between your local agent and the cloud.
Remote Agent Runtime
Each cloud server starts with a short-lived, scoped agent credential and an MCP proxy. The remote agent can write code, mount datasets, and request compute without any AWS access.
Simple by Default
Tell your local agent what you need: “Upload these FASTQ files, start a cloud server, run STAR, and show me the results.” Pepkio turns that request into infrastructure and a remote agent ready to execute.
Secure by Design
Agents never receive raw AWS credentials. Remote agents get short-lived, scoped tokens. Sensitive actions can require explicit human approval.
Scientific Data Locality
Keep large research datasets and compute in the same cloud region, with optimized uploads for large files. Public reference datasets can be mounted read-only on the cloud server.
Elastic Compute
Your remote agent requests the resources it needs. Pepkio provisions or resizes the cloud server and scales it down when it is no longer needed.
After your cloud server is running
Who does what once analysis is underway — not the setup steps above.
“Upload these FASTQ files, start a cloud server, run STAR, and show me the results.”
plans the work, creates the project and cloud server, uploads data, and selects tools.
provisions the cloud server and injects a scoped remote agent runtime.
writes code, mounts reference data, requests compute if needed, and runs the analysis.
the analysis — while your local agent monitors progress and handles errors.
“Upload these FASTQ files, start a cloud server, run STAR, and show me the results.”
plans the work, creates the project and cloud server, uploads data, and selects tools.
provisions the cloud server and injects a scoped remote agent runtime.
writes code, mounts reference data, requests compute if needed, and runs the analysis.
the analysis — while your local agent monitors progress and handles errors.
Human in the Loop. Agents in Control.
Agents should move fast. But they shouldn't accidentally create an expensive bill or delete research data.
Your remote agent works on the cloud server and can ask for more compute or mount shared reference data. Pepkio pauses high-risk actions and asks for your approval:
Autonomous workflow approval gate
Permission Request
Request a 64GB workspace. Estimated cost: $42/day.
Start in Seconds
With Pepkio Cloud, you don't spend time:
Built With Early Users
Pepkio Cloud is currently in Early Access. We're working directly with researchers to support real-world bioinformatics workflows, add requested datasets and software environments, and improve the platform based on hands-on use.
Have a dataset, reference genome, tool, or workflow you need? Tell us what you're trying to run. We'll help you get started.
Common Questions
- How do I get started?
- Sign in to the Cloud console and create a project, or connect Cursor / Claude Code / Codex as a local agent. Full step-by-step instructions are in the getting started guide. After your cloud server is running, connect a remote agent via SSH to run analysis on it.
- Do I need an AWS account or AWS credentials?
- No. Pepkio Cloud provisions and manages cloud servers, IAM permissions, S3 buckets, and networking automatically. Agents never receive raw AWS credentials.
- How does Human-in-the-Loop work for safety?
- High-risk actions such as a remote agent requesting a high-memory cloud server, exceeding project spend caps, or deleting original datasets pause execution and request your approval in the browser before proceeding.
- Where is research data stored and processed?
- Datasets and compute run in the same cloud region for high throughput and scientific data locality, with optimized chunked uploads for large genomics and transcriptomics files. Public reference datasets can be mounted read-only at /datasets/... on the cloud server.
- Which tools and bioinformatic pipelines are supported?
- Your remote agent can run standard tools (STAR, BWA, samtools, DeepVariant, Nextflow) or custom Python/R scripts directly on isolated, elastic Linux cloud servers. It can use the shell and filesystem like a local developer, while your local agent handles cloud operations.