Veliora is a multi-agent biomedical research system for drug repurposing that funds workflows in USDC, runs evidence-driven pipelines, and delivers structured, traceable research briefs.
Built for the Agentic Economy on Arc Hackathon
- Per-step paid research actions
- Pricing tied to actual execution
- Payments are triggered dynamically during the workflow
- Economic activity is driven by task progression
- The PI agent coordinates paid actions across the workflow
- Machine-to-machine payment flows follow task progression
- Live app: https://veliora.vercel.app/
- Video: Watch on YouTube
- Slides: veliora.pdf
Veliora is a multi-agent biomedical research system that turns a disease-focused query into a structured repurposing workflow.
A user funds a research job in USDC, and the PI agent coordinates specialist agents across literature mining, drug database screening, pathway analysis, hypothesis generation, evidence scoring, conditional critique, report synthesis, and review.
The result is a structured research brief that may surface reportable candidates, exploratory hypotheses, or pipeline-reviewed signals, depending on the strength of the run.
Arc provides the escrow and settlement layer, while x402 and Circle Gateway enable paid access to external services through on-demand payment flows without interrupting the core agent orchestration.
This turns fragmented research steps into a coordinated, economically structured workflow, where agent-to-agent actions follow task progression and outputs remain traceable.
Biomedical research workflows are composed of many specialized steps:
- literature retrieval and filtering
- drug and target screening
- pathway and biology anchoring
- evidence evaluation and critique
These steps are difficult to coordinate economically.
Traditional payment systems:
- make research actions too expensive,
- bundle work into opaque services,
- and provide limited visibility into multi-stage workflows.
This makes it hard to build modular, auditable, and fairly priced research pipelines.
Veliora combines an agentic research workflow with a multi-layer payment system:
- ERC-8183 on Arc → job escrow, lifecycle, and resolution
- x402 + Circle Gateway → per-step paid research/review actions
- Arc → deterministic sub-second finality with USDC-denominated fees and settlement
This enables:
- escrowed research jobs
- conditional paid research and review actions
- traceable workflow execution
- peer-reviewed delivery
- rejection + refund for non-deliverable outputs
Veliora turns fragmented research steps into a coordinated, economically viable system.
- Escrowed research jobs via ERC-8183
- Paid per-step research actions via x402 + Circle Gateway
- Multi-agent workflow with specialized roles
- Review-gated delivery and refund-on-rejection
- Traceable evidence pipeline from input to output
A user submits a disease-focused request, such as:
- Idiopathic pulmonary fibrosis
- Type 2 Diabetes
- Multiple sclerosis
- Alzheimer’s disease
- Parkinson’s disease
- Triple-negative breast cancer
“What repurposable compounds may be relevant for this condition based on literature, pathway context, and known targets?”
The system:
- creates and funds a job in USDC
- dispatches specialist agents
- executes paid research steps
- gathers and scores candidate evidence
- challenges reportable candidates when a scored shortlist is available
- assembles the final research brief and review outcome
- delivers a shortlist, surfaces exploratory signals, or rejects the run
- Strong signal → report delivered
- Weak signal → exploratory report
- No scored shortlist, but usable early leads → pipeline-reviewed signals surfaced as exploratory context
- No defensible result → rejected + refunded
Veliora operates across three layers:
-
Client Layer
Job creation, USDC approval, and escrow funding via ERC-8183 on Arc -
Execution Layer
PI agent orchestrates the funded workflow across literature, DrugDB, pathway, repurposing, evidence, red-team critique, report synthesis, and peer review. -
Resolution Layer
Evaluator completes or rejects the job and triggers escrow payout or refund
sequenceDiagram
participant U as Client Wallet
participant J as ERC-8183 Job on Arc
participant P as PI Agent
participant S as Paid Research Actions
participant E as Evaluator
U->>J: createJob
P->>J: setBudget
U->>J: approve + fund
J->>P: job funded
P->>S: request paid research action
S-->>P: 402 Payment Required
P->>S: replay with payment proof
S-->>P: verified response
P->>E: submit for evaluation
E-->>P: approve or reject
alt approved
E->>J: complete
else rejected
E->>J: reject
end
Paid actions cover literature, DrugDB, pathway, red-team critique, and peer review, while repurposing, evidence scoring, and report synthesis remain in the PI-led core orchestration layer.
| Agent | Responsibility |
|---|---|
| PI Agent | Orchestrates the workflow, manages the escrow state, and dispatches research steps |
| Literature | Retrieves and filters disease-relevant papers from the literature layer |
| DrugDB | Screens molecules, targets, and activity-linked candidate context |
| Pathway | Anchors disease biology through pathway, genetic, and clinical trial context |
| Repurposing | Generates candidate hypotheses from upstream evidence |
| Evidence | Scores and prioritizes shortlisted candidate signals |
| Red Team | Challenges shortlisted candidates for risk and confounders |
| Report | Synthesizes findings into a structured research brief |
| Reviewer I | Verifies methodology, provenance, and report completeness |
| Reviewer II | Checks consistency across candidate, evidence, and scoring fields |
| Tiebreaker | Resolves disagreements when reviewers reach different conclusions |
-
Literature
Uses PubMed / NCBI for paper retrieval, PubMed Central when full-text artifacts are available, and OpenAlex for citation enrichment and ranking context. -
DrugDB
Uses ChEMBL for molecule, target, and activity-linked candidate context. -
Pathway
Uses Open Targets plus linked pathway, genetic, and active clinical-trial context, including ClinicalTrials.gov references where available.
Veliora separates payments into three layers:
- Client funds job in USDC
- Evaluator completes or rejects
- Approved jobs trigger escrow payout
- Rejected jobs are refunded
Used during execution for:
- literature retrieval
- DrugDB queries
- pathway analysis
- read team critique
- peer review
Flow:
- request resource
- receive
x402 Payment Required - sign authorization
- replay with payment
- seller verifies payment and serves the resource
- Gateway batches authorizations and settles them on Arc
Price: 0.002 USDC per action
- Triggered after successful completion
- Distributed across based on contribution
- Not executed for rejected jobs
The funded workflow is documented and supported by a batch-level transaction summary covering escrow funding, seller settlement evidence, run-level execution, and the micropayment ledger.
Full transaction summary: artifacts/hackathon-batch/transaction-summary.md
ArcScan references for ERC-8183 role activity:
-
Provider / PI wallet (
0x871bf9e2a99de1bd47a2a0c4d1056349eb4ab8a8) —setBudgetandsubmittransactions https://testnet.arcscan.app/address/0x871bF9E2A99DE1bD47a2a0c4D1056349eB4AB8a8?tab=txs -
Evaluator wallet (
0x99f89acf1a0db9665452a8fa5b50c43757e320a1) —completetransactions https://testnet.arcscan.app/address/0x99f89acf1a0dB9665452A8fA5b50c43757E320A1?tab=txs
- Next.js
- TypeScript
- Wagmi
- Viem
- Node.js
- Express
- SQLite
- Vercel
- Railway
- Arc Deterministic-finality L1 with USDC-denominated fees and settlement
- ERC-8183 Job escrow, submission, and evaluator-based resolution lifecycle
- USDC Funding and settlement asset used across job escrow and paid workflow actions
- Circle Wallets
Developer-controlled wallets for role-based wallet operations - Circle Gateway
Gasless authorization and batched nanopayment settlement - x402
Payment-gated access to research/review actions
- OpenRouter LLM inference for agent stages
- Custom multi-agent pipeline Project-specific orchestration across the end-to-end research workflow
Veliora evaluates outputs across:
- literature support
- biological relevance
- clinical evidence
- safety profile
- genetic context
Outputs are research prioritization artifacts
Full rubric: REPORT_QUALITY_RUBRIC.md
Veliora is intentionally selective.
- strong shortlist exists
- exploratory hypotheses or pipeline-reviewed signals that did not advance into the final shortlist
Pipeline-reviewed signals
Early mechanism-linked leads surfaced by the literature, drug, pathway, and candidate-review workflow. These are exploratory and are not the same as a final shortlist.
Red-team challenged candidates
Stronger candidates that reached the shortlist stage and were stress-tested by the red-team layer before delivery.
- no defensible signal
- review fails
Rejected jobs are refunded onchain.
Veliora is a payment-aware, multi-agent biomedical research pipeline.
It combines:
- ERC-8183 for escrowed jobs
- x402 + Circle Gateway for per-step payments
- Arc for fast finality and USDC-native settlement
The result is a system that can execute, evaluate, and economically structure complex research workflows with selective, review-gated outputs.