中文Open DemoStart a Pilot
PERSLIS SCIENCE / API + MCP

Keep your model.
Add Lois.

When an answer needs evidence, route it through Lois by API or MCP. She applies your source policy, builds Peel’s proprietary evidence trace, tests the claim, and returns structured evidence to the application you already use.

See the architecture Talk to Lois ↗
API + MCPKeep your model
THE LOIS EVIDENCE LAYER

Any model in. An evidence-governed result out.

Lois is the Perslis model your application calls when an answer needs evidence, provenance, hypothesis testing, or trusted-source constraints. She does not replace the model you already use. She gives it a governed scientific reasoning layer.

YOUR STACKExisting AI application
GPTClaudeGeminiLlamaYour model
API
or
MCP
PERSLIS SCIENCE
L

Lois

Evidence-constrained scientific reasoning

  1. Apply source policy
  2. Search + aggregate
  3. Build Peel trace
  4. Test the claim
STRUCTURED
RESULT
BACK TO YOUR APPEvidence result
ClaimsSourcesContradictionsUnknownsTrace
THE PROPOSITION

Keep your model. Add Lois.

Use the language model, agent framework, research product, or proprietary ML system you already trust. Route only the questions that need evidence through Lois.

WHAT YOU KEEP

Your model

Language, ambiguity, planning, extraction, presentation, and the user experience you already built.

GPTClaudeGeminiLlamaML
WHAT YOU ADD

Lois

Permitted-source retrieval, evidence standing, contradiction checks, visible unknowns, and a complete provenance chain.

NO SOURCE → NO CLAIM
ONE MODEL · SIX OPERATING MODES

Choose how Lois should reason.

The calling application sets the job and the source boundary. Lois returns the result, its standing, and the path that produced it.

01trusted_source

Trusted Source Mode

Restrict Lois to PubMed, FDA, named journals, internal data, or any allowlist you control.

02evidence

Evidence Mode

Return a claim only when it connects to evidence. Unsupported statements remain exposed.

03hypothesis

Hypothesis Mode

Generate candidate explanations while separating support, contradiction, and the unknown for each one.

04discovery

Discovery Mode

Search across the permitted corpus for relationships the original question did not anticipate.

05verify

Verify Mode

Submit an answer from another model and ask Lois: can this claim survive the evidence?

06provenance

Provenance Mode

Return the complete chain for citations, evidence graphs, audit trails, and Peel traces.

API + MCP

A contract your application can inspect.

Send an objective, a mode, the material to evaluate, and a source policy. Receive evidence as structured state—not prose you are forced to trust.

01
Request LoisIllustrative API contract
POST
{
  "objective": "Can this claim survive the evidence?",
  "mode": "verify",
  "input": {
    "claim": "BRCA1 loss predicts PARP inhibitor response"
  },
  "source_policy": {
    "allow": ["PubMed", "FDA", "internal:trial-results"]
  }
}
02
Structured evidence resultRendered however your product needs
JSON
{
  "standing": "partially_supported",
  "claims": [{ "status": "supported", "evidence": 7 }],
  "contradictions": [{ "evidence": 1 }],
  "unknowns": ["clinical context not specified"],
  "provenance": ["PMID:…", "FDA:…", "internal:…"],
  "peel_trace": { "inspectable": true, "retained": true }
}
TRACE
Peel's proprietary evidence trace

Each inspectable trace can carry provenance, source identity, claim standing, contradictions, unresolved unknowns, relationships, hypotheses, and the checks performed—so another system can render citations, graphs, audit trails, or a full research record.

Schema shown for product illustration. Confirm the supported production contract, source connectors, quotas, and deployment requirements during your API pilot.

ONE EVIDENCE MODEL · MANY PRODUCTS

Lois goes where the question happens.

Perslis Science becomes infrastructure other products consume—not another destination researchers must remember to visit.

AI

AI products

Give a chat or copilot an evidence path for the moments when fluent generation is not enough.

AG

Agents

Let an agent call Lois before it promotes a researched claim into a plan or report.

RS

Research software

Add source-governed answers and complete traces inside the workspace scientists already use.

EN

Enterprise systems

Combine approved public sources with internal company data without hiding where a conclusion came from.

THE GOVERNING RULE

Models propose. Lois tests. The evidence decides.

  • Source policy firstThe caller controls which evidence Lois may use.
  • Unknown is validMissing support remains a gap instead of becoming generated filler.
  • Your product presents itLois returns structured state; your application controls the final experience.
ADD LOIS TO YOUR STACK

Give any model an evidence layer.

Start with one high-value question, one source policy, and one product surface. We will scope the API or MCP path with you.

Start an API pilotTalk to Lois
SEE IT IN ACTION

Explore the connected science workspace.

A recorded tour of the scientific tools and research workspace.

Recorded on a research prototype. The recording shows the scope demonstrated at that time.

Explore the research floor yourself →
PERSLIS / SCIENCERECORDED DEMO
A recorded tour of the scientific tools and research workspace.
ONE CONTINUOUS THREAD

From a question to a record you can inspect.

01Ask
02Retrieve
03Pin
04Verify
05Retain
See how it works →
BRING LOIS A REAL QUESTION

Let your next discovery start here.

Start with a target, a dataset or a research question. Inspect what the evidence supports, what it rejects and what remains unknown.

Start a research pilot →

Everything Perslis