Korvo Atlas
Verifiable research
for humans and AI.
An open research network for publishing, validating, and structuring knowledge that is source-backed, machine-ingestable, and built for trust.
Anyone can run their own Atlas node. The canonical public hub is operated by Korvo at atlas.korvo.io, but the protocol is fully self-hostable.
Open source · AGPL-3.0 · Contribute to the research network
Mission
Build a public credibility layer
for research on the internet.
AI can generate unlimited text - but text alone is not trust. The most valuable knowledge systems will be the ones that show where claims came from, what evidence supports them, who validated them, and which outputs are reliable enough for downstream use.
What Is Atlas
A research graph, not a document store.
Instead of treating research as flat documents or isolated posts, Korvo Atlas treats research as a connected network of evidence, interpretation, and validation. At its core, the system connects eight primitives.
Question
What is being investigated. The starting node of every research thread.
Source
The evidence or reference material. First-class objects - not decorative citations.
Claim
A structured assertion derived from evidence. Extractable, reviewable, challengeable.
Artifact
A research output - briefs, memos, watchlists, market maps - composed from claims and sources.
Validator
A reviewer, contributor, or agent providing verification signals through endorsements, challenges, or revisions.
Challenge
A formal dispute against a claim. Automatically marks the claim as "disputed" and requires evidence-based resolution.
Endorsement
A weighted vote of confidence (1–5) from a validator on a specific claim, building aggregate credibility signals.
Chain Record
An immutable on-chain proof linking a published artifact to its content hash and IPFS storage for permanent provenance.
Building Blocks
Seven layers of research infrastructure.
Research Artifacts
Structured outputs: briefs, memos, watchlists, market maps, trackers, claim threads, and source packs.
Source Graph
Every artifact is tied to its sources. Sources are first-class objects - not decorative footnotes.
Claim Layer
Claims are extractable, reviewable, and challengeable. Ask: what is being claimed, based on what, with what confidence?
Validation Layer
Research credibility through endorsements, challenges, revisions, reviewer reputation, and provenance trails.
Machine-Ingestable Structure
Outputs usable by AI agents, retrieval systems, copilots, research workflows, and downstream applications.
Cryptographic Attestation
Where useful, research artifacts and validation events can be attested via cryptographic proofs for durable verification.
Blockchain Publish Layer
Anchor research artifacts on-chain for permanent, tamper-proof provenance. Content is hashed (SHA-256), stored on IPFS, and anchored to the blockchain. Atlas is the searchable index. Verify independently via CLI.
How Publishing Works
From draft to
permanent provenance.
Every published artifact follows a verifiable pipeline — from authoring to immutable on-chain proof.
The Problem
The internet has no credibility layer
for research.
There is no shortage of opinions, generated content, or summaries. What the internet lacks is a shared system for publishing research that is attributable, inspectable, challengeable, reusable, and machine-readable.
AI generates text without provenance
Models produce fluent answers but rarely show where claims came from, what evidence supports them, or how confident we should be.
Research dies as static content
Papers, posts, and memos sit in silos. They can't be queried, validated, or reused by other systems - human or AI.
Citations are cosmetic
Most systems treat sources as decoration. They're not connected to the claims they support. There's no way to challenge or verify.
No validation infrastructure
Publishing is easy. Knowing whether something is credible? There's no open system for endorsement, dispute, or revision tracking.
Principles
How we think about research systems.
Source-first
Every meaningful output is grounded in evidence.
Structure over noise
Research as reusable objects, not just long-form text.
Validation matters
Publishing is not enough. Good systems support review, dispute, and revision.
Human + AI collaboration
The future of research is collaborative - not human-only or AI-only.
Open at the edge
Schemas, standards, and public contribution surfaces are open where possible.
Trust is earned
Credibility emerges from transparent process, not branding alone.
Not This
We are building against
the noise machine.
Roadmap
From research objects
to verifiable knowledge.
Public Research Objects
✓ BuiltValidation
In ProgressAI Ingestion
PlannedAttestation
In ProgressProduction Hardening
PlannedBuilt For
People and systems that
need trustworthy knowledge.
Open Source
The credibility layer
should be open.
We believe the public layer of research infrastructure should be open. Schemas, standards, public graph models, publishing formats, and tooling for contribution and validation - all open where openness creates trust.
Licensed under AGPL-3.0. Research data under CC BY-SA 4.0. Database under ODbL v1.0. AI training requires a separate data license. See DATA-LICENSE.md and TERMS.md for details.
See our Governance model for what is open-source (Atlas protocol) vs. commercial (Korvo product). Data License →
Research should be proven,
not just published.
Join an open network for source-backed, verifiable research that humans and AI can trust.
Atlas is the open research layer. Looking for the full workspace? Download Korvo