Getting started with Tracer

Tracer is the layer your authorization or onboarding system calls before a transaction goes through. It runs your fraud, risk, and limit policies in milliseconds and returns ALLOW, DENY, or REVIEW. The decision then lives in one place, not scattered across product code. What changes in your operation: decision logic stops living in scattered if statements across services. Rule changes ship through an API the same day, not in the next release. Validation history gives you one place to investigate why a transaction received its decision. Trade-off to be honest about: you add one HTTP call to the critical path of every transaction. The target is p99 under 80ms. In return, you get a single point for policy and decision history, and you remove duplicated logic from product code.

Who is this guide for? Developers (junior or senior) integrating Tracer for the first time. If you’re evaluating Tracer at a product or strategy level, start with What is Tracer. If you already have it running and need API mechanics, jump to the Tracer API quick start.

This guide walks you through setting up Tracer and running your first validation. In a few steps, you’ll have a working environment ready to validate transactions in real time. For step-by-step API instructions with request and response examples, see the Tracer API quick start.

Why use Tracer

Real-time validation: Make ALLOW/DENY/REVIEW decisions in under 80ms (p99)
Flexible rules: Expression-based rule engine for custom business logic
Spending control: Configure limits by account, portfolio, segment, and period
Validation history: Stored decisions for investigation and reporting
Product-agnostic: Supports any transaction type (Card, Wire, Pix, Crypto)

By the end of this guide, you will:

Understand Tracer architecture and core concepts
Have a working development environment
Run your first transaction validation
Configure a spending limit

What is Tracer

Tracer is a transaction validation platform that evaluates rules and limits and returns instant decisions. Your system calls Tracer before it executes a transaction. It then acts on the decision (ALLOW, DENY, or REVIEW) according to your business logic.

How it works

How Tracer processes a validation request across its Validation, Rules, and Limits contexts and returns an ALLOW, DENY, or REVIEW decision — Figure 1. How Tracer works

In this flow:

Rules evaluate expressions against the transaction context
Limits check spending thresholds for applicable scopes
Decision returns ALLOW, DENY, or REVIEW based on evaluation results

Core contexts

Tracer has three bounded contexts:

Validation Context - Orchestrates requests, coordinates evaluation, and stores validation history
Rules Context - Manages rule definitions and expression evaluation
Limits Context - Manages spending limits and usage tracking

Prerequisites

Before you start, make sure you have:

Docker and Docker Compose installed
Go 1.26+ for local development (the repository’s go.mod declares the exact toolchain version)
PostgreSQL 17 (the shared Midaz primary, started by the platform infrastructure compose, not by Tracer’s own)
API Key for authentication

Infrastructure dependencies

Tracer requires the following components:

Component	Version	Purpose
PostgreSQL	17	Data persistence. Tracer uses its own `tracer` database on the shared Midaz PostgreSQL primary; it does not ship a dedicated instance.

Ports

Default ports used by Tracer services:

Service	Port	Description
Tracer API	4020	Main REST API
PostgreSQL	5701	Shared Midaz PostgreSQL primary, as exposed by the shipped infrastructure example (`DB_PORT`)

Step 1: Set up the environment

You can run Tracer with Docker Compose or locally for development.

Option A: Docker Compose (recommended)

Tracer’s own Compose file declares only two services: the application and a one-shot migration runner. PostgreSQL is not one of them. It comes from the shared platform infrastructure Compose and must be healthy first. The application container starts only after the migration runner applies the schema and exits successfully. The service always boots against an already-migrated database.

In Midaz v4, Tracer is source-available under ELv2 in the Midaz repository and release. It still runs as its own service. Start from components/tracer for local development.

Navigate to the Tracer project directory and start the services:

cd components/tracer

# Setup environment
cp .env.example .env

# Start all services (brings up the shared infrastructure first,
# then the migration runner, then Tracer)
make up

Option B: Local run

For development, you can run Tracer locally:

# Set environment variables
export DB_HOST="localhost"
export DB_NAME="tracer"
export API_KEY="your-secure-api-key"
export API_KEY_ENABLED="true"
export SERVER_PORT="4020"
export LOG_LEVEL="INFO"

# Start the service
go run cmd/app/main.go

Essential environment variables

Variable	Description	Example
`DB_HOST`	PostgreSQL host	`localhost`
`DB_NAME`	PostgreSQL database name	`tracer`
`API_KEY`	API Key for authentication	`your-secure-api-key`
`API_KEY_ENABLED`	Enable API Key authentication	`true`, `false`
`SERVER_PORT`	API port	`4020`
`LOG_LEVEL`	Log level	`INFO`, `DEBUG`, `ERROR`

Step 2: Authenticate to the API

Tracer supports API key and plugin authentication. Plugin authentication takes precedence when you enable both, except on endpoints configured as API-key-only.

Deployment and authentication configuration	Auth header	When to use
Single-tenant, plugin auth enabled	`Authorization` header with a Bearer token	`MULTI_TENANT_ENABLED=false` and `PLUGIN_AUTH_ENABLED=true`, unless the endpoint is configured as API-key-only.
Single-tenant, API key auth	`X-API-Key` header	`MULTI_TENANT_ENABLED=false` and plugin auth is disabled, or the endpoint is configured as API-key-only.
Multi-tenant (SaaS / BYOC Multi-Tenant)	`Authorization` header with a Bearer token	Any deployment with `MULTI_TENANT_ENABLED=true`. Plugin auth is required; the JWT is issued by Access Manager and carries the `tenantId` claim.

The remaining steps use the single-tenant API key form because most local-development setups run that way. If plugin authentication applies to your request, replace X-API-Key: your-secure-api-key with Authorization: Bearer $JWT in every example.

API Key (single-tenant)

Include the API Key in the X-API-Key header:

GET /v1/rules
X-API-Key: your-secure-api-key

Bearer JWT (multi-tenant)

Include the JWT issued by Access Manager in the Authorization header:

GET /v1/rules
Authorization: Bearer eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9...

Tracer extracts the tenantId claim from the JWT and routes the request to the correct tenant database. You never pass the tenant identifier in a header, path, body, or rule scope. The token is the only source of truth.

cURL example

# Single-tenant: List rules
curl -H "X-API-Key: your-secure-api-key" \
  http://localhost:4020/v1/rules

# Multi-tenant: List rules
curl -H "Authorization: Bearer $JWT" \
  https://tracer.sandbox.lerian.net/v1/rules

Keep API Keys and JWTs secure. Never expose them in client-side code or public repositories.

API key authentication is disabled by default (API_KEY_ENABLED=false). The provided .env.example keeps it off, so local development works without setup. A production deployment must set API_KEY_ENABLED=true (single-tenant) or MULTI_TENANT_ENABLED=true and PLUGIN_AUTH_ENABLED=true (multi-tenant) before you expose the service.

Step 3: Configure a spending limit

Spending limits control transaction amounts by scope and period. Create a limit using POST /v1/limits.

Limit types

Type	Description	Period counting
`DAILY`	Maximum amount per day	A new count starts each calendar day at 00:00 UTC
`WEEKLY`	Maximum amount per week	A new count starts each ISO week, Monday at 00:00 UTC
`MONTHLY`	Maximum amount per month	A new count starts on the 1st of the month at 00:00 UTC
`CUSTOM`	Maximum amount for a custom date range	One count for the whole range
`PER_TRANSACTION`	Maximum per single transaction	No count is kept

For detailed configuration of all limit types including time windows and custom periods, see the Spending limits guide.

Scopes

Apply limits to specific contexts:

Segment: Apply to all accounts in a segment (e.g., corporate customers)
Portfolio: Apply to accounts in a portfolio
Account: Apply to a specific account
Transaction type: Apply only to CARD, WIRE, PIX, or CRYPTO

Create a limit

curl -X POST http://localhost:4020/v1/limits \
  -H "X-API-Key: your-secure-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Daily Corporate Card Limit",
    "description": "Daily spending limit for corporate card transactions",
    "limitType": "DAILY",
    "maxAmount": "50000.00",
    "asset": "BRL",
    "scopes": [
      {
        "segmentId": "550e8400-e29b-41d4-a716-446655440000",
        "transactionType": "CARD"
      }
    ]
  }'

Activate a limit

curl -X POST http://localhost:4020/v1/limits/{id}/activate \
  -H "X-API-Key: your-secure-api-key"

Limit lifecycle

Limits start in DRAFT status and follow the lifecycle DRAFT → ACTIVE → INACTIVE. Inactive limits can return to DRAFT for editing, or you can delete them permanently. Activate a limit to start enforcement. For the full lifecycle and transition rules, see the Spending limits guide.

Monitor usage

Every POST /v1/validations response carries limitUsageDetails, with one entry per limit Tracer checked. Each entry holds the cap, the amount attempted, and the projected consumption of that cap’s current period. That projection includes this transaction. The endpoint GET /v1/limits/{id}/usage reports a cumulative total across the limit’s counters, for a review of overall consumption. For detailed configuration options, see the Spending limits guide.

Step 4: Validate your first transaction

With limits configured, you’re ready to validate a transaction using POST /v1/validations.

Submit a transaction for validation

Send a validation request with the transaction context including:

Transaction details (type, amount, asset, timestamp)
Account information
Optional: segment, portfolio, merchant, and custom

TS=$(date -u +%Y-%m-%dT%H:%M:%SZ)

curl -X POST http://localhost:4020/v1/validations \
  -H "X-API-Key: your-secure-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "requestId": "550e8400-e29b-41d4-a716-446655440104",
    "transactionType": "CARD",
    "subType": "credit",
    "amount": "1500.00",
    "asset": "BRL",
    "transactionTimestamp": "'"$TS"'",
    "account": {
      "accountId": "550e8400-e29b-41d4-a716-446655440100"
    },
    "merchant": {
      "merchantId": "550e8400-e29b-41d4-a716-446655440103",
      "category": "5411",
      "name": "Test Merchant"
    },
    "metadata": {
      "channel": "mobile"
    }
  }'

The transactionTimestamp must be recent, which is why the example generates it. Tracer rejects a future timestamp with error code 0419 (1-minute clock skew tolerance). Tracer rejects a timestamp older than 24 hours with error code 0421.

requestId is the idempotency key. Send a new UUID for each attempt. If you repeat one, Tracer returns the decision it already recorded for that key. A rule you activated in between will not appear to take effect.

Tracer evaluates the rules and limits that apply to the transaction, then returns one of three decisions:

Decision	Meaning	Your system should
`ALLOW`	Transaction approved	Proceed with the transaction
`DENY`	Transaction denied (rule matched or limit exceeded)	Block the transaction
`REVIEW`	Requires manual review	Queue for human review

The response names the rules Tracer evaluated, the rules that matched, and the current limit usage. That detail helps with debugging and customer support.

Why Tracer returns a decision instead of blocking directly. Tracer works as a decisioning layer, not an authorization gateway. The calling system holds the customer relationship and knows the channel. It decides what to do with a DENY. For example, your card-issuing system may honor a DENY for a stand-in pre-auth. It may still capture the request for analytics. By returning a decision, Tracer fits into any authorization flow without owning the customer-facing UX.

For complete payload structure and field details, see the API reference.

Step 5: Create a validation rule

Rules let you define custom business logic that evaluates during validation. Create a rule using the POST /v1/rules endpoint with an expression, action, and optional scopes. For example, to block high-value transactions:

curl -X POST http://localhost:4020/v1/rules \
  -H "X-API-Key: your-secure-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Block high-value card transactions",
    "description": "Deny card transactions above R$ 10,000",
    "expression": "amount > 10000",
    "action": "DENY",
    "scopes": [
      {
        "transactionType": "CARD"
      }
    ]
  }'

Tracer preserves the name’s case and internal whitespace, trimming only leading and trailing whitespace before storage. Take the ruleId from the response and use it in the activate call below.

Activate a rule

curl -X POST http://localhost:4020/v1/rules/{id}/activate \
  -H "X-API-Key: your-secure-api-key"

Activation takes effect right away on the instance that served the activate call. A single-instance setup evaluates the rule on your next validation. When you run several instances behind a load balancer, the others pick the change up on their next rule sync. The interval is RULE_SYNC_POLL_INTERVAL_SECONDS, default 10. Deactivation propagates the same way.

Rule lifecycle

Rules follow the same lifecycle as limits: DRAFT → ACTIVE → INACTIVE. To start evaluation, activate the rule using POST /v1/rules/{id}/activate. You can deactivate and reactivate an active rule as needed. For detailed information about rule expressions and lifecycle management, see the Rules engine guide.

Observability

Tracer exposes endpoints for monitoring and observability.

Key metrics

Tracer exposes OpenTelemetry-compatible metrics via the OTLP exporter, plus custom application metrics:

tracer_auth_failures_total{reason} - Authentication failures by reason (missing_api_key, invalid_api_key)
tracer_validation_rollback_failures_total - Usage rollback failures during REVIEW decisions (eventual consistency gaps that self-correct at period boundaries)

Tracer’s built-in OpenTelemetry HTTP middleware provides standard HTTP request metrics automatically.

Verification

Confirm that everything works correctly.

Checklist

Docker services started and healthy
API Key authentication working
Spending limit configured
Test transaction validated successfully
Rule created and activated

Next steps

You’ve set up Tracer and validated your first transaction. From here, you can explore more advanced features:

Integration guide - Learn how to integrate your authorization system with Tracer
Rules engine - Write validation rules in CEL and manage their lifecycle
Spending limits - Configure and manage spending limits by scope and period
Validation history and compliance - Query validation history and use it in your compliance processes

Quick reference

The three flows you’ll use most:

Validate a transaction: POST /v1/validations. See the Tracer API quick start for the request shape.
Manage rules: /v1/rules (CRUD + lifecycle endpoints /activate, /deactivate, /draft). See the Rules engine guide.
Manage limits: /v1/limits (CRUD + lifecycle + /usage). See the Spending limits guide.

For the full endpoint catalog, request/response schemas, and error codes, see the API reference.

What your system should do with each decision

Decision	Tracer recommends	Your system should
`ALLOW`	Approval	Proceed with the transaction
`DENY`	Denial	Block the transaction and inform the user
`REVIEW`	Review	Queue for manual review in your system

Tracer returns decisions as recommendations. Your system must implement the appropriate action for each decision.

Overview

Guides

Console

Operations

Getting started with Tracer

Why use Tracer

What is Tracer

How it works

Core contexts

Prerequisites

Infrastructure dependencies

Ports

Step 1: Set up the environment

Option A: Docker Compose (recommended)

Option B: Local run

Essential environment variables

Step 2: Authenticate to the API

API Key (single-tenant)

Bearer JWT (multi-tenant)

cURL example

Step 3: Configure a spending limit

Limit types

Scopes

Create a limit

Activate a limit

Limit lifecycle

Monitor usage

Step 4: Validate your first transaction

Submit a transaction for validation

Step 5: Create a validation rule

Activate a rule

Rule lifecycle

Observability

Key metrics

Verification

Checklist

Next steps

Quick reference

What your system should do with each decision

​Why use Tracer

​What is Tracer

​How it works

​Core contexts

​Prerequisites

​Infrastructure dependencies

​Ports

​Step 1: Set up the environment

​Option A: Docker Compose (recommended)

​Option B: Local run

​Essential environment variables

​Step 2: Authenticate to the API

​API Key (single-tenant)

​Bearer JWT (multi-tenant)

​cURL example

​Step 3: Configure a spending limit

​Limit types

​Scopes

​Create a limit

​Activate a limit

​Limit lifecycle

​Monitor usage

​Step 4: Validate your first transaction

​Submit a transaction for validation

​Step 5: Create a validation rule

​Activate a rule

​Rule lifecycle

​Observability

​Key metrics

​Verification

​Checklist

​Next steps

​Quick reference

​What your system should do with each decision

Why use Tracer

What is Tracer

How it works

Core contexts

Prerequisites

Infrastructure dependencies

Ports

Step 1: Set up the environment

Option A: Docker Compose (recommended)

Option B: Local run

Essential environment variables

Step 2: Authenticate to the API

API Key (single-tenant)

Bearer JWT (multi-tenant)

cURL example

Step 3: Configure a spending limit

Limit types

Scopes

Create a limit

Activate a limit

Limit lifecycle

Monitor usage

Step 4: Validate your first transaction

Submit a transaction for validation

Step 5: Create a validation rule

Activate a rule

Rule lifecycle

Observability

Key metrics

Verification

Checklist

Next steps

Quick reference

What your system should do with each decision