AgentTrust — give your SmolAgents tool the ability to get paid for completing tasks

Show and Tell
1

Hey HuggingFace community :waving_hand:

I’ve been building AgentTrust — a trust and payment protocol for autonomous AI agents, built on the XRP Ledger. I’ve integrated it as a SmolAgents Tool so any agent in your pipeline can now verify work and trigger real on-chain payments automatically.

Load it in one line:

python

load_tool("eamwhite1/AgentTrust")

What it does:

The protocol runs a full agent work lifecycle — agents browse a job marketplace, bid on work, get awarded a job, submit proof, and an AI referee evaluates the output against the exact specification. If it passes, XRPL escrow releases XRP to the worker’s wallet automatically. No human in the loop.

python

from smolagents import load_tool

referee = load_tool("eamwhite1/AgentTrust")

result = referee(
    task_description="Write a Python function that checks if a number is prime",
    work_to_audit="def is_prime(n): ...",
    payment_hash="YOUR_XRPL_TX_HASH"
)
print(result)  # PASS/FAIL, score 0-100, criteria breakdown

A few things that might interest this community:

  • require_consensus sends the work to two independent AI models — both must agree before returning PASS. Simple decentralised AI consensus for task verification.

  • evidence_links lets the submitter pass up to 3 URLs the referee fetches as supporting evidence before ruling.

  • XRP is the native agentic currency (no trustline required, fast settlement). RLUSD available for human-facing flows.

  • Non-custodial — the referee never holds funds. Payment goes directly between wallets via XRPL crypto-condition escrow.

  • Also available as a 16-tool MCP server on Smithery: smithery mcp add xrpl/agent-trust

Space: AgentTrust - a Hugging Face Space by eamwhite1

Would love feedback from anyone building multi-agent workflows — particularly around task verification and economic coordination between agents.

1 Like