Touchstone is a drop-in MCP server. Add it to any MCP client — Claude Desktop, Claude Code, any host — and your agent gets four tools: rank x402 services by their on-chain trust trace, and pay through the atomic gate while keeping its own signing key. Touchstone only proposes; the agent verifies and signs.
Add it — one mcpServers entry
{
"mcpServers": {
"touchstone": {
"command": "python",
"args": [".../mcp_server.py"],
"env": { "ALGOD_URL": "https://your-algorand-node" }
}
}
}
Four tools your agent gets
A touchstone is a slab of black stone. Rub gold across it and the color of the trace reveals its purity — instantly, with no paperwork and no one's word to take. We do exactly that for services: rub each candidate against Algorand's permanent record, and read the truth off the trace.
A drop-in MCP server — no migration, no marketplace to join. Your agent calls four tools and gets back signed evidence it checks itself, while keeping its own signing key. Touchstone only proposes; the agent verifies and signs.
We compute the trace off-chain, then hand the power to Algorand. The check rides inside the payment as one atomic group — so there's no seam, no instant where money moves but the check doesn't.
Your agent needs a service. Touchstone pulls live x402 listings from the GoPlausible discovery API — real Algorand endpoints, today — and fetches each candidate's on-chain payment history, the gold to be tested.
Is the demand real, or wash-traded to fake popularity? The chain is the stone; the trace is the trust score — published with the limits of what it can't catch.
Your rule decides — pay only if the trace beats your bar. Payment, rule-check and record settle as one indivisible group. If it would fail, nothing moves.
The receipt is now a permanent, public fact — and it feeds the next test. Every honest call darkens the record and sharpens the stone.
A faker can pay itself to look popular — that's wash trading. Counting payments can't catch it: a fake ring and a real hit both look busy. So we don't count. We ask: could innocent luck explain this timing? Shuffle the real history thousands of times, then measure how far the true pattern sits from that innocent cloud. That distance is the trace.
QuickPin's self-paying ring lands at +7.5 — miles outside anything luck produces. Caught by timing. Atlas, a real service, sits at −0.8, right inside the innocent range.
Catches bursts and coordination. But a patient faker can spread payments thin to look natural. MapRush did — timing alone scores it −2.3, a miss.
Follow the money backwards. If one wallet quietly seeded every "customer," it shows no matter how the timing is spread. MapRush's provenance: 0.77 — caught.
In a blind test on synthetic wash/honest traffic the detector scored 0% false alarms, 100% precision. Sub-floor and max-effort fakes are reported as not certified, never "caught" — a bounded, falsifiable claim, still to be validated on a real on-chain split.
Plenty of projects now score agents. Almost all produce a number you check on a server and then choose to obey. Touchstone is the only one that puts the check inside the payment, over a trace you can recompute yourself.
Reflects each project's public materials as of mid-2026 and its design approach — not a head-to-head benchmark.
A rule is just the limits Touchstone enforces for you. Use a preset or build your own, then give each agent a rule and watch which sample it would pay — and the trace that sample leaves on the stone.
illustrative — synthetic samples chosen to show how the rules behave; the trace colors mirror the real wash-detector verdict.
The presets
How pure the trace must be before you'll pay.
Reject if fakes could hide below this share of volume.
The most you'll spend on a single request.
Planned: require a stake that's slashed if the evidence proves false.
Four samples on the stone
Your agents — tap to reassign a rule
Drop an agent to Adventurous and it pays a base-metal sample — the stone still showed the dull trace, but a low bar lets it through. The rule is yours; the chain enforces exactly the one you set.
Suppose something forced your agent to pay QuickPin — the base-metal sample — even though it fails the bar.
Because the trace-check rides inside the payment, Algorand refuses the whole transaction. It reverts on-chain — not a cent moves. The agent is protected even from its own mistake.
Verified on Algorand — in the live agent run, 5 agents: 3 settled, 2 protected (one reverted by the gate on-chain, one declined at discovery; a low-bar agent settled a wash by its own choice).
Touchstone settles in USDC through GoPlausible, the live Algorand x402 facilitator. The same gate rides straight to Mainnet. Here's what's already real.
loading on-chain proof…
Pulled live from the facilitator's discovery API — a neutral registry, not yet Touchstone-scored. Settlements shown are facilitator-wide (all operators), not Touchstone volume.