Legiti

v0.2.0-alpha.1API Keyalpha

BR ticketing-native fraud prevention (formerly Inspetor, São Paulo). 4th fraud entry. Order evaluation + Collection API for events/accounts/auth tracking. Alpha on npm — public docs thinner than ClearSale/Konduto; tool shapes match docs-backend repo.

18 tools2 env vars🇧🇷 Brazilalpha — pending validation

Install

terminal

$ npm install @codespar/mcp-legiti@alpha

Quickstart

Add this entry to your claude_desktop_config.json (or any MCP-compatible client config).

claude_desktop_config.json

{
  "mcpServers": {
    "legiti": {
      "command": "npx",
      "args": [
        "-y",
        "@codespar/mcp-legiti@alpha"
      ],
      "env": {
        "LEGITI_API_KEY": "<your_legiti_api_key>"
      }
    }
  }
}

After restart, your agent can call any of the 18 tools below — try evaluate_order first.

Environment variables · 2

Required to authenticate the server. Stored encrypted when using CodeSpar managed hosting.

LEGITI_API_KEY

requiredsecret

Legiti API key (JWT-format bearer token). Use the sandbox key while testing — Legiti trains its ML models on every request, so test data hitting a production key pollutes the model.

LEGITI_BASE_URL

optional

Legiti API base URL. Defaults to https://collection-prod.inspcdn.net. Legiti issues customer-specific base URLs — override per contract.

Available tools · 18

Each tool is independent — your agent loads only what it needs to reduce context and response time.

READevaluate_order

Submit an order to Legiti for real-time fraud evaluation via the v2 order endpoint. Returns a decision (approve / reject / manual) synchronously — response may take up to ~20s. Include as much context as possible: account, payment (tokenized card bin+last4), CPF, billing/shipping, and for ticketing flows the event_date_id / sale_items.

WRITEupdate_order

Notify Legiti of a status change on an existing order (e.g. payment captured, shipped, cancelled, refunded). Feeds the ML model — required for ongoing decision quality. Use mark_order_fraudulent for confirmed chargebacks instead.

READmark_order_fraudulent

Report a confirmed chargeback / fraud outcome back to Legiti. This is Legiti's primary ML feedback channel — unreported chargebacks degrade future decision quality for similar buyers. Call this after the issuer confirms the chargeback, not on mere suspicion.

READevaluate_sale

Legacy single-shot sale evaluation via POST /evaluation. Synchronous — returns { inspetor_decision: 'approve' | 'reject' | 'manual' }. Prefer evaluate_order (v2) for new integrations; use this when you only have the minimal required fields or for feature parity with older Legiti/Inspetor integrations.

WRITEtrack_account

Notify Legiti of an account lifecycle event (created / updated / deleted). Legiti's ML model treats the Account as a primary entity and needs every state change to score future sales accurately.

WRITEtrack_event

Notify Legiti of an Event (concert, show, match, session) lifecycle change. Events are primary entities in Legiti's ticketing-native model — scoring for ticket sales depends on up-to-date event metadata (date, venue, capacity, price tiers).

WRITEtrack_sale

Notify Legiti of a Sale state change (created / updated). For initial sale creation without asking for a decision, set request_evaluation=false on evaluate_order instead. Use this for post-creation updates that aren't status transitions (e.g. sale items added/removed before capture).

WRITEtrack_auth

Notify Legiti of an authentication or password event (login attempt, logout, password recovery request, password reset). Login/logout and password activity are strong signals for account-takeover fraud — feed every attempt, successful or failed.

Compatible frameworks

MCP is a protocol — any framework that speaks it can mount this server.

Claude Agent SDKAnthropic

OpenAI AgentsOpenAI

Vercel AI SDKVercel

LangChainLangChain

LlamaIndexLlamaIndex

CrewAICrewAI

MastraMastra

Any MCP clientProtocol

Frequently asked questions

All credentials are encrypted at rest in an isolated vault scoped per organization. OAuth2 tokens are automatically refreshed before expiry. Your agent code never sees raw secrets — it receives scoped, short-lived session tokens.