Five layers. Your science comes in at the bottom, is held as connected evidence, and is read by agents that write what they find back. The layer in the middle is Attest, and it is the reason the rest works.
01
Connectors
Your science arrives where it already lives — notebooks, assays, instrument output, the thread where someone said the run was never repeated. Nothing has to be migrated first.
ELN / LIMS
Lab results & experiment logs
Chats & threads
Literature
02
Attest · the evidence layer
Every claim is stored with the source it came from, what supports it, what conflicts with it, and what depends on it. This is the layer the other four are built on.
AttestDB · open source
Claims & frames
Provenance on every write
Conflicts & gaps
Cascade retraction
03
Compute
Analyses that actually run, in notebooks and pipelines you can open. Your own discovery models score candidates on what your organization knows.
Notebooks & pipelines
Compute routing
Your discovery models
04
Agents
AI scientists that read the evidence layer, run the work, and write results back to it. Build them, share them, reuse them.
Literature investigation
Data analysis
Hypothesis & experiment design
05
Where you work
The same answers in the tools your scientists already use. Adoption does not require anyone to move.
via Omic MCP
Omic Discover
Claude & Claude Code
ChatGPT
Layer 02 · every claim one click from its source. Open full size ↗
Attest, in the middle
A search tool finds information. It has no opinion about whether what it found is still true, whether another team contradicted it last year, or what else would break if it were wrong. Attest holds those as facts of their own.
01
What it stores
A claim, the passage it came from, and the confidence it carries. Not a document in an index — a position, with its basis attached.
02
What it connects
Supporting evidence, conflicting evidence, and the gap where a result was never followed up. The buried negative is visible because the shape around it is.
03
What it protects
Retract a source and every conclusion built on it is flagged, wherever it ended up. That is cascade retraction, and it is why a wrong answer does not quietly survive.
AttestDB, the store underneath, is open source. Your team can read how provenance and retraction work rather than take our word for it.
An answer with its sources and conflicts, an analysis you can rerun, the experiment that would settle the question, and a record the next person starts from. Follow one question through it →
Layer 03 · the notebook and its figures, open and rerunnable. Open full size ↗
