Loop, then network.

Cryptographic commit/reveal, evidence grading rubric, three-persona red team — all running locally.

• · sha-256 commit on every mission
• · Evidence grades A/B/C/D/X with explicit rubric
• · Skeptic / Scientist / Senior Reviewer veto rules

Literature, sequence/structure, target/pathway, variants, ADMET, novelty, patent, grader, thesis, red team, synthesizer, dossier.

• · Real PubMed / UniProt / OpenTargets / ChEMBL / AlphaFold calls
• · Pool-level tier classification (upstream / reasoning / output)
• · Per-task budget tracking

Editorial dashboard with live SSE updates and replayable mission cards.

• · 12 dashboard panels
• · SSE event stream
• · Buyer-safe dossier preview

Live peptide synthesis pricing across vendors, attached to dossier as commitable cost prediction.

• · 3+ vendor quotes per dossier
• · Commit synthesis cost in mission hash
• · Auto-flag when scope drifts

Auto-draft SPR/BLI binding and serum-stability protocols, version-controlled, hashed.

• · Protocol DSL (yaml-like)
• · Per-assay prediction commits
• · Replay against assay readout

Submit measured readouts; Pepclaw auto-grades the prediction vs. result and updates trust score.

• · Per-mission Brier score
• · Public verifiability
• · Falsified predictions are first-class artifacts

pepclaw verify <mission-id> — recompute the commit hash from any machine.

• · Standalone Go binary
• · Reads only public commit + salt
• · Reproduces evidence index

Third parties can register agent pools that Pepclaw routes to (paid per task, evaluated by trust score).

• · Pool registration manifest
• · Per-pool reputation history
• · Slashing for fabricated PMIDs

Pharma buyers subscribe to dossier streams matching their target classes.

• · Per-target subscription
• · Auditable replay packs
• · On-chain settlement (optional)