Peptide research,
replayable by anyone.
Pepclaw is an autonomous research swarm for nonclinical peptide discovery. Twelve specialist agent pools query the world's biomedical databases, grade their own evidence, and ship a buyer-safe dossier — every claim hashed before the run, revealed after, and verifiable forever.
Tamper-evident by design.
Pepclaw doesn’t try to make biology trustless. It makes the claims, predictions, artifacts and state transitions tamper-evident, replayable and economically accountable.
- Step 01Commit
Each mission's question and target are hashed with a per-mission salt before any agent runs. Pepclaw cannot change the question post-hoc.
- Step 02Ingest
Upstream pools pull live evidence from PubMed, UniProt, AlphaFold, OpenTargets, ChEMBL and Reactome. Every finding carries source provenance.
- Step 03Reason
Reasoning pools score novelty, grade evidence A through X, and compose falsifiable theses with explicit mechanism claims.
- Step 04Critique
A three-persona red team — Skeptic, Scientist, Senior Reviewer — challenges every thesis. Only the Senior may issue a hard block.
- Step 05Reveal
On completion, Pepclaw publishes the salt. Anyone can re-hash the original question and verify the run was honest end-to-end.
One orchestrator. Twelve specialists.
Upstream — data ingestion
Real biomedical APIs · provenance preserved- Literature Minerliterature_miner
PubMed E-utilities. Pulls peer-reviewed papers with full PMID provenance and structured abstracts.
- Sequence & Structuresequence_structure
UniProt + PDB + AlphaFold. Annotates peptide targets, structural confidence, lipidation tolerance.
- Target & Pathwaytarget_pathway
OpenTargets GraphQL + Reactome. Maps targets to diseases and mechanism-of-action pathways.
- Variant Linkervariant_linker
ChEMBL REST. Links targets to known small molecules and binding-assay endpoints.
- ADMET Developabilityadmet_developability
Heuristic ADMET scorecards: solubility, half-life, peptide developability, off-target risk.
Reasoning — pattern & critique
Whitespace, grading, hypothesis, red team- Novelty Scoutnovelty_scout
Whitespace detector. Scores each finding for novelty against historical mission corpus.
- Patent Competitivepatent_competitive
Patent landscape and freedom-to-operate surveillance via Lens.org-compatible signals.
- Thesis Generatorthesis_generator
Composes structured, falsifiable hypotheses with evidence-cited mechanism claims.
- Evidence Graderevidence_grader
Grades each finding A/B/C/D/X using study type, replication and reporting strength.
- Red Teamred_team
Three personas — Skeptic / Scientist / Senior Reviewer. Only Senior can issue a hard block.
Output — synthesis & dossier
Buyer-safe markdown · PMID-cited · hedged- Synthesizersynthesizer
Consolidates cross-pool findings and theses into the synthesis document.
- Dossier Assemblerdossier_assembler
Buyer-safe markdown dossier with PMID citations, hedged claims and full evidence chain.
No mocks.
No fabricated PMIDs.
Every finding Pepclaw produces is anchored in a real source identifier you can click. Every PubMed citation has a real PMID. Every UniProt entry resolves. Every OpenTargets association is recomputable. If a finding cannot be traced to its source, it cannot enter the dossier.
- PubMedlive
- UniProtlive
- AlphaFoldlive
- OpenTargetslive
- ChEMBLlive
- Reactomelive
- Lens.org Patentspending
- ClinicalTrials.govfuture
Built so you don’t have to trust us.
Falsifiability is a first-class artifact.
Every mission commits a sha-256 hash before it runs and reveals the salt after. Auditors can replay the loop forever.
Evidence is graded, not gestured at.
Findings are scored A through X using study type, replication and reporting strength — and the rubric ships in /docs.
The red team can actually block.
Three personas critique each thesis. Only the Senior Reviewer holds veto, and only when mechanism is unfalsifiable or evidence is fabricated.
A clean falsification is a successful run.
The only unacceptable outcome is an uninterpretable result. Negative answers ship as dossiers too.