AI ToolDead Man Walking

ESM Atlas Review

ESM Metagenomic Atlas is an open atlas of billions of predicted metagenomic protein structures, enabling biological discovery and protein design.

shipped Jun 1, 2026aifreemium

Read full review↓

Visit ESM Atlas↗

airesearchproduct-hunt

1ESM Atlas provides access to 617 million predicted metagenomic protein structures.

2The database contains 1.1 billion predicted protein structures and 6.8 billion protein sequence entries.

3Launched on May 27, 2026, by the biomedical research institute Biohub.

4Structures were generated by the AI model ESMFold2, which is fully open-source and allows unrestricted commercial use.

𝕏 in ↑↗

Stork Quadrant

Dead Man Walking· 15/100

An LLM can do most of what this tool's UI promises. No moat, no agent presence.

“The 617 million predicted metagenomic protein structures are the only thing keeping this alive. No LLM can conjure that corpus from thin air — it's a specific, curated, computationally expensive dataset that took Meta's ESM model and massive infrastructure to produce. The UI is replaceable; the atlas is not. But it's a single moat, and Meta owns it, so any defensibility belongs to them, not a downstream wrapper.”
— Claude Sonnet 4.6, scored 2026-06-01

Defensibility · 15/100

Physical-world coupling
Regulatory moat
Network liquidity
Proprietary refreshing data
High-trust catastrophic workflows
Multi-party coordination
Brand / community / taste

An LLM alone could replace

Explain what a protein structure looks like or describe its properties in natural language
Summarize research papers about metagenomic proteins
Generate hypotheses about protein function based on sequence descriptions
Answer general questions about metagenomics and protein folding concepts

Agent-Readiness · 15/100

Verified MCP
Listed on agent surfaces
Usage-based pricing
Headless agent auth
Public OpenAPI— https://esmatlas.com/openapi.json
Active changelog
llms.txt— https://esmatlas.com/llms.txt

How to defend

The only real move is to become the query and analysis layer that researchers actually cite — build tooling for structural comparison, functional annotation pipelines, and integration with wet-lab workflows so the atlas becomes infrastructure, not just a search box.

Ship an MCP server and list it on Stork — biggest single point gain (+25).
Get listed in the Anthropic MCP registry, Cursor, or Claude Desktop (+20).
Add a usage-based or per-call tier; per-seat-only pricing dies when agents replace seats (+15).
Expose API-key auth with a self-serve sandbox tier; remove sales-call gates (+15).
Publish a public changelog and ship in the last 90 days — silence reads as abandonment (+10).

How this score is computed →See the full quadrant How to defend

ESM Atlas at a Glance

Best For

Researchers and developers in metagenomics and bioinformatics

Pricing

freemium

Key Features

Open access to metagenomic protein structures, Comprehensive database of predicted structures, User-friendly interface for researchers, Supports various research applications, Regular updates with new data

Alternatives

AlphaFold Protein Structure Database, RoseTTAFold (Baker Lab), OpenProtein.AI, OmegaFold

About ESM Atlas

Platforms

Web

Target Audience

Researchers and developers in metagenomics and bioinformatics

Leadership

Meta AI

Connect

𝕏

X / Twittertwitter.com/esmatlas

</>Embed "Featured on Stork" Badge▼

HTML

<a href="https://www.stork.ai/en/esm-atlas" target="_blank" rel="noopener noreferrer"><img src="https://www.stork.ai/api/badge/esm-atlas?style=dark" alt="ESM Atlas - Featured on Stork.ai" height="36" /></a>

Markdown

[![ESM Atlas - Featured on Stork.ai](https://www.stork.ai/api/badge/esm-atlas?style=dark)](https://www.stork.ai/en/esm-atlas)

overview

What is ESM Atlas?

ESM Atlas is a structural biology AI tool developed by Biohub (founded by Mark Zuckerberg) that enables researchers in metagenomics and structural biology to access and explore predicted metagenomic protein structures. It provides access to 617 million predicted metagenomic protein structures and was generated by the AI model ESMFold2. This open atlas serves as a critical resource for understanding protein function, facilitating drug discovery, and exploring uncharacterized biological diversity. The underlying ESMFold2 model predicts 3D protein structures directly from amino acid sequences and was released alongside ESMC, a state-of-the-art protein language model.

quick facts

Quick Facts

Attribute	Value
Developer	Biohub (Meta AI Founders)
Business Model	Freemium (Open Access Core)
Pricing	Free (with API rate limits)
Platforms	Web, API
API Available	Yes
Founded	May 27, 2026
Target Audience	Biologists, Bioinformaticians, Structural Biologists, Researchers in protein science, Disease researchers

features

Key Features of ESM Atlas

ESM Atlas offers a robust set of features designed to support advanced research in structural biology and metagenomics. Its core functionality revolves around providing extensive access to predicted protein structures and the tools to leverage this data.

1Access to 617 million predicted metagenomic protein structures.
2Comprehensive database featuring 1.1 billion predicted protein structures and 6.8 billion protein sequence entries.
3Structures generated by the advanced AI model ESMFold2, which predicts 3D protein structures from amino acid sequences.
4ESMFold2 is fully open-source and allows unrestricted commercial use, promoting broad scientific application.
5An API is available at https://esmatlas.com/about#api for programmatic access to the database, subject to rate limits.
6Organizes proteins by relationships learned by the model, revealing novel connections not found in traditional databases.
7Supports protein design capabilities, enabling researchers to design and test new functional proteins.
8Incorporates a large amount of microbial protein data, offering insights into diverse environmental biology.
9User-friendly interface designed for researchers in various biological fields.

use cases

Who Should Use ESM Atlas?

ESM Atlas is primarily designed for the scientific community, providing a foundational resource for various research and development activities in biology and medicine.

1Biologists: For understanding protein function, exploring uncharacterized biology, and inferring roles of proteins in biological processes.
2Bioinformaticians: For advanced bioinformatics applications, data analysis in genomics, and integrating large-scale structural data into computational workflows.
3Structural Biologists: For protein structure prediction, exploring metagenomic protein diversity, and analyzing protein shapes to inform experimental design.
4Researchers in protein science: For biological discovery, protein design, and accelerating early therapeutic binder discovery against targets in cancer and immunology.
5Disease researchers: For designing new drugs and therapeutics, and exploring protein connections relevant to poorly understood diseases.

pricing

ESM Atlas Pricing & Plans

ESM Atlas operates on a freemium model, providing open access to its comprehensive database of predicted metagenomic protein structures. The underlying ESMFold2 model is fully open-source and permits unrestricted commercial use, a significant departure from some proprietary models. Access to the API is available, though it is subject to rate limits on sequence length and the number of requests per user to ensure fair usage as a shared resource. Specific paid tiers for increased API capacity or enterprise features are not publicly detailed, but the core resource is freely accessible for research and development.

1Open Access: Free access to the 617 million predicted metagenomic protein structures via the web interface.
2API Access: Free with rate limits on sequence length and number of requests per user, as it is a shared resource.
3ESMFold2 Model: Fully open-source and available for unrestricted commercial use, allowing local deployment and custom applications.

competitors

ESM Atlas vs Competitors

ESM Atlas and its underlying ESMFold2 model are positioned as a significant advancement in structural biology, directly challenging existing solutions with its scale, performance, and open-source nature.

AlphaFold Protein Structure Database↗

Offers a vast, highly accurate database of over 200 million predicted protein structures, covering nearly all catalogued proteins known to science.

Similar to ESM Atlas in providing a large, open-access database of predicted protein structures for research. While ESM Atlas specifically focuses on metagenomic proteins and emphasizes speed with its language model, AlphaFold is renowned for its high accuracy across a broader range of proteins and has a larger overall database size, though not exclusively metagenomic.

RoseTTAFold (Baker Lab)↗

Integrates deep learning with traditional energy-based methods to predict tertiary protein structures and protein-protein interactions, including complete biological assemblies.

Unlike ESM Atlas, which is a pre-computed atlas of metagenomic structures, RoseTTAFold is a powerful AI prediction tool that researchers use to generate structures on demand, including protein complexes, rather than browsing a pre-existing database.

OpenProtein.AI↗

Provides a no-code platform with powerful foundation models for protein engineering, structure/function prediction, and model training, making advanced AI accessible to biologists.

While ESM Atlas is a static atlas of predicted structures, OpenProtein.AI offers an interactive platform for designing and predicting new proteins using AI, including custom model training. It targets researchers but focuses on active protein engineering rather than just providing access to a pre-computed database, and offers a free tier for academia.

OmegaFoldOn Stork Compare

A single-sequence based model that excels at predicting structures for orphan proteins and in antibody design without requiring multiple sequence alignments (MSAs), offering a balance between speed and accuracy.

Similar to ESMFold (the underlying model for ESM Atlas) in being a single-sequence based prediction tool, OmegaFold offers an alternative for researchers needing fast predictions, especially for proteins lacking evolutionary information. Unlike the pre-computed ESM Atlas, OmegaFold is a tool for on-demand prediction, often used for novel or de novo designed proteins.

❓

Frequently Asked Questions

+What is ESM Atlas?

+Is ESM Atlas free?

Yes, ESM Atlas operates on a freemium model, providing open access to its comprehensive database of predicted metagenomic protein structures via its web interface. API access is also free but is subject to rate limits on sequence length and the number of requests per user. The underlying ESMFold2 model is fully open-source and allows unrestricted commercial use.

+What are the main features of ESM Atlas?

Key features include access to 617 million predicted metagenomic protein structures, a database of 1.1 billion predicted protein structures generated by the ESMFold2 AI model, an open-source license for ESMFold2 allowing commercial use, an available API for programmatic access, and the ability to organize proteins by learned relationships to surface novel biological connections.

+Who should use ESM Atlas?

ESM Atlas is intended for biologists, bioinformaticians, structural biologists, and researchers in protein science and disease research. It supports understanding protein function, biological discovery, protein design, exploring metagenomic protein diversity, and drug discovery applications.

+How does ESM Atlas compare to alternatives?

ESM Atlas differentiates itself from alternatives like AlphaFold by offering a larger database of 1.1 billion predicted structures, a specific focus on metagenomic proteins, and an underlying ESMFold2 model that is fully open-source for commercial use and claims superior performance in protein complex prediction. Unlike prediction tools such as RoseTTAFold or OmegaFold, ESM Atlas primarily serves as a pre-computed atlas, while OpenProtein.AI focuses on interactive protein engineering.

For builders

This page is doing a job for someone else’s tool.

AI agents read it. Buyers find it. Backlinks accrue. Your tool can have one too — live in 24 hours, indexed by Claude, ChatGPT, and Perplexity, queryable via MCP.

List your tool What you get