The Verdict API
You don't want to be the judge.
Name us in your terms once.
Recourse is a Verdict API: dispute bundle in — terms, interaction log, deliverable — reasoned ruling out in minutes. Fault in percentages, a money instruction your escrow partner executes, and the reasoning published behind every decision. The same engine that powers the dispute room, at platform scale.
Quickstart
Your first verdict in five minutes
# 1. Sign up at /signup → generate a key (25 free rulings, no card)
# 2. POST your first dispute:
curl https://api.recourse.so/disputes \
-H "authorization: Bearer rk_live_..." \
-H "content-type: application/json" \
-d '{ "rulepack": "marketplace-contractor-v1", "currency": "USD", "amount_minor": 60000,
"terms": { "format": "text", "content": "Logo package, one revision round, 10 days, $600." },
"claim": { "by": "claimant", "claimed_breach": "Not professional quality." },
"log": [ { "at": "2026-06-12T16:30:00Z", "actor": "respondent", "kind": "delivery", "content": "Delivered per spec." } ],
"deliverables": [] }'
# → fault allocation, money instruction, reasoning trace. That's the integration.Or skip the terminal entirely: the playground runs the same engine in your browser.
The problem you own today
Every platform ends up running an amateur court
It's expensive
Support agents with macros deciding whether a logo "met the spec" cost $5–15 a ticket fully loaded — before the make-goods and close-the-ticket refunds they hand out to end conversations. Recourse rules at $4 list, in minutes.
It's inconsistent
Outcomes vary by agent, workload, and time of day — and your users compare notes. Recourse decides every case under the same published doctrine, with the reasoning attached, so like cases get like outcomes.
It's conflicted
The venue taking a cut of the transaction is deciding who keeps the money. Every ruling you issue makes your brand the defendant. "The independent venue decided" is a different conversation entirely.
And in the EU, it's now regulated
DSA Article 21 obliges platforms to engage with out-of-court dispute settlement — at human-body prices, case by case. Recourse is built to be the AI-native body ruling at machine cost (our Article 21 analysis).
Who plugs in
Three integrations, one endpoint
Marketplaces
Freelance, services & P2P marketplaces
Your order terms, message thread, and deliverable already form the evidence bundle — POST /disputes takes the queue as-is. Draft verdicts land in your agents' Zendesk sidebar during the pilot; rulings release funds through your PSP. Worked example: the missed trade show →
Escrow & payment rails
Escrow providers & PSPs
Escrow without adjudication is a timelock — "what happens when the parties disagree" is the missing sentence in your docs. One signed webhook completes the product: executable on fiat escrow, Stripe Connect, Tempo, or x402, with our liability behind the ruling while you keep custody and the customer. The lifecycle →
Vertical SaaS
Property, bookings, recruitment & trades software
If your software sits between landlords and tenants, clients and contractors, or agencies and candidates, their disputes already arrive in your support inbox — without you ever agreeing to judge them. Name the venue in your terms and hand the judgment to a neutral. Talk to us about your vertical →
Adjacent
Agents & independent professionals
Building for AI agents transacting with each other? That's its own integration and its own page. Serving independent professionals with no platform in the middle? The email witness brings the venue to a plain email thread.
Measured, not trusted
The 30-day shadow pilot
You shouldn't outsource judgment on a pitch. A shadow pilot runs Recourse silently beside your existing process on your live queue: your team decides as they do today, the engine rules the same cases in parallel, and nothing changes for your users. After 30 days you get the report — agreement rate with your best agents, time-to-ruling, cost per case, and every divergence with both reasonings side by side — and you decide with numbers. We get measured; you risk nothing.
Set up a shadow pilot Typical setup: an export of anonymized closed tickets, or the Zendesk app on a test view.
The venue goes where the deal is
Summon the court from anywhere
The API is the engine — but disputes start in tickets, inboxes, and group chats, so the venue meets them there. Every channel is a front door to the same doctrine and the same reasoned ruling:
⌘ Verdict API live ✉ CC it on the email thread early access ◳ Zendesk sidebar pilot ✈ Telegram group bot waitlist ◎ WhatsApp group waitlist ⬡ MCP server for agents in dev
All channels, statuses, and waitlists → Channels
The lifecycle
How it works
Submit the bundle
POST /disputes with the agreed terms, the chronological log, and the deliverables. Your systems already have all three.
Get the ruling
Percentage fault allocation under published doctrine, a money instruction that sums exactly to the disputed amount, a rule-cited reasoning trace, and a confidence score. Minutes, not weeks.
Enforcement executes
One signed webhook to your escrow partner: release, refund, or split. Recourse never touches funds.
Escalation when warranted
Messy evidence, high values, or low confidence route up the ladder to Recourse's own review bench — never back to your team. See below.
Every tier is ours, never yours
Who judges? The escalation ladder
Every tier is staffed by Recourse, never by your platform. That is the point of the product: a platform that judges its own transactions is conflicted — the judge profits from the deals it rules on. When you name Recourse in your terms, judgment leaves your building entirely.
| Tier | Who decides | When | SLA |
|---|---|---|---|
| Tier 1 — Automated verdict | The Recourse judgment engine, ruling under published doctrine | Clean evidence bundles — the bulk of any queue | Minutes |
| Tier 2 — Enhanced review | A deeper engine pass, confirmed or overridden by a Recourse reviewer on our internal bench | Messy or contradictory evidence, engine confidence below threshold, amounts above the automatic ceiling | Same business day |
| Tier 3 — Human panel | Independent adjudicators contracted by Recourse — trained on the doctrine, conflict-screened per case, excluded from any dispute where they have a relationship with either party or the platform | High-value disputes, fraud allegations, novel doctrinal questions | 3 business days |
Until a Tier 2/3 review concludes, the escrow instruction ships flagged escalated: true and your escrow partner holds funds. The reviewer's ruling supersedes it with a final signed instruction. Platforms pay for the tier used; platforms never influence the outcome — our fee is never a percentage of the disputed value, so the venue has no stake in who wins.
Published rate card — never a share of the dispute
Pricing
Free trial
Your first 25 rulings free, any tier. No card required until they're used. Plus the playground — no account at all.
Start freePay as you go
No minimum, no commitment. $20.00 enhanced review · $150.00 human panel. Card on file after the trial; the month's usage is charged on the 1st. How billing works.
Get a keyVolume & enterprise
Annual volume commitments with reserved capacity, liability-backed rulings (insurance-capped warranty), DSA/ODS readiness, shadow-pilot onboarding, named-in-terms support.
Talk to usThe platform pays, never the disputants. Never a percentage of the disputed value — a court's fee cannot depend on the judgment. Anonymised-rulings data-rights discount available on committed plans.
Wear the venue
The badge your users can trust
Customers who name Recourse in their terms display the badge on checkouts, listings, and proposals. It tells both sides of every transaction the same thing: if this goes wrong, an independent venue decides — not the platform, not the counterparty.
Licensed to accounts in good standing where the venue is actually named — so the badge never lies.
Badge assets & terms of use
