Full feature reference

Everything your stack
needs from a read layer

From real-time CDC to auto-scaling — every capability ReflexDB ships with, explained in detail.

Sync

Real-time CDC sync

ReflexDB propagates row-level changes to the in-memory snapshot in under a second under normal load. No polling, no batch jobs, no stale reads.

MySQL / MariaDB — binlog CDC via native replication protocol
PostgreSQL — logical replication CDC via WAL decoding
SQL Server — CDC change table polling (coming soon)
Position persisted across restarts (binlog offset / LSN) — no full reload on reconnect
Exponential backoff reconnect on stream loss
Automatic fallback to poll mode if CDC is unavailable
On-demand reload via POST /reload
Read-replica support — point at a replica to avoid load on primary

Poll mode is available on all plans (minimum 300 s on Free, 5 s on paid). CDC sync modes (binlog, logical, cdc) are available on Pro and above.

sync log

# Binlog CDC stream
connected  host=db.prod.internal port=3306
position   file=mysql-bin.003821 pos=94021

INSERT orders id=10482 status="pending"
UPDATE orders id=10480 status="shipped"
UPDATE orders id=10481 status="shipped"
DELETE orders id=10477

snapshot   swapped in 0.4 ms  rows=184,201
lag        220 ms behind source

API

REST API generated from your schema

At build time, ReflexDB reads your database schema and generates a typed server with one endpoint per table. No code to write — declare your config and deploy.

Filter on any column: =, !=, <, >, IN, LIKE, ILIKE, IS NULL
Nested filter references — filter a parent by a condition on its children
FK relation traversal — embed related rows in a single request
Aggregations: COUNT, SUM, AVG with implicit GROUP BY
ORDER BY (multi-field), LIMIT, OFFSET
OpenAPI 3 spec + interactive API explorer at GET /
Column aliasing, field notes, and per-table row predicates via YAML config

reflexdb.yaml → HTTP

# reflexdb.yaml
tables:
  orders:
    fields: [id, status, total_cents, created_at]
    relations:
      customer: customers.id

# Query with ReflexQL
POST /query HTTP/1.1
Content-Type: text/plain

orders(status = "shipped") {
  id status total_cents created_at
  customer { id name }
} ORDER BY created_at DESC LIMIT 20

Compute

Right-sized instances with clean multipliers

Pick the instance size that fits your dataset. Compute units use clean power-of-2 multipliers — easy to reason about how many units a workload consumes.

5 sizes from nano (0.25 vCPU / 0.5 GB) to 2xlarge (4 vCPU / 16 GB) — larger instances available by request on Business+
Each size maps to dedicated compute — pick what fits your dataset, ReflexDB manages the rest
Compute auto-selected based on your schema size at build time
Overage billed per compute unit at $25/unit — no surprise tiers

instance sizes

size      vCPU   RAM     units/mo
────────────────────────────────────
nano      0.25   0.5 GB  0.85
small     0.25   1 GB    1.00  ← 1 unit
medium    0.50   2 GB    2.00
large     1.00   4 GB    4.00
xlarge    2.00   8 GB    8.00
2xlarge   4.00   16 GB   16.00

# Pro plan: 4 units included
# Example: 2× medium = 4 units exactly

Infrastructure

Deploy instances close to your users

Choose a region per instance. ReflexDB routes all operations to the correct region automatically — one dashboard, multiple deployment targets.

Ireland and Virginia available today
Choose a region when creating each instance
Isolated infrastructure per region — your data never shares compute with other customers
Instance images deployed to all active regions at build time
Available on all plans

See the regions page for current availability and the rollout roadmap.

management API

# Set default region on a connection
PATCH /connections/conn_01J… HTTP/1.1

{
  "defaultRegion": "us-east-1 (Virginia)"
}

# Override per instance at creation
POST /databases HTTP/1.1

{
  "connectionId": "conn_01J…",
  "region": "eu-west-1 (Ireland)",
  "instanceSize": "large"
}

Security

Connect private databases inside a VPC

Your production database doesn't need to be publicly accessible. ReflexDB supports SSH tunnel connections — configure a jump host and the tunnel is established automatically.

SSH jump host with private key authentication
Private key stored in an encrypted secrets store
Pre-flight connection test before every build and rebuild
API key authentication on every generated endpoint
Your data is never logged, stored, or accessible to ReflexDB staff
GDPR-compliant: Europe (Ireland) default; data residency guaranteed

SSH tunnel config

# Connection with SSH tunnel
POST /connections HTTP/1.1

{
  "host": "10.0.1.15",
  "port": 3306,
  "username": "reflexdb_ro",
  "ssh": {
    "host": "bastion.internal.example.com",
    "port": 22,
    "username": "ec2-user",
    "privateKey": "-----BEGIN RSA..."
  }
}

# Pre-flight test runs on every build
status: testing_connection → queued

Scaling

Auto-scaling and auto-pause

Enterprise plans include horizontal auto-scaling — multiple replicas behind a load balancer that scale based on CPU utilisation. The Free tier pauses after two days of inactivity to keep shared infrastructure costs manageable.

Auto-scaling with horizontal replicas (Enterprise)
Query traffic automatically distributed across replicas
Free tier auto-pauses after 2 days of inactivity
Resume with any request — cold start is under 30 seconds

Auto-pause only applies to the Free tier. Paid plans run 24/7 with no automatic pausing.

instance status

# Instance running — 2 replicas
GET /databases/db_01J… HTTP/1.1

{
  "status": "running",
  "instanceSize": "large",
  "region": "eu-west-1 (Ireland)",
  "runningReplicas": 2,
  "maxReplicas": 2,
  "endpoint": "https://db_01J….reflexdb.cloud"
}

Developer platform

Full control plane via REST API

Every action in the dashboard is available via a documented REST API. Automate provisioning, rebuilds, and monitoring in your CI/CD pipeline.

Long-lived management API keys (rmk_… prefix) scoped to your account
OpenAPI 3 spec at GET /openapi.json — import into any HTTP client
Interactive API reference at docs.reflexdb.cloud
Provision, rebuild, terminate, and inspect instances programmatically
Webhook subscriptions for instance lifecycle events
Usage endpoint: GET /account/usage returns build credits and compute units consumed

management API

# Trigger a rebuild via API key
POST /databases/db_01J…/rebuild HTTP/1.1
Authorization: Bearer rmk_01J…

{ "reason": "schema change" }

# Poll build status
GET /databases/db_01J…/builds HTTP/1.1

{
  "builds": [{
    "status": "running",
    "startedAt": "2026-05-30T14:22:01Z",
    "computeType": "small"
  }]
}

AI & MCP

Built-in MCP server for AI agents

Every ReflexDB instance includes a Model Context Protocol server that exposes your data as queryable tools. Connect Claude, Cursor, or any MCP-compatible client and let AI agents query your live database — no glue code, no extra infrastructure.

MCP server auto-generated from your instance schema — no configuration required
Query live data from Claude, Cursor, Windsurf, and other MCP-compatible tools
CDC sync means AI agents always see the current state of your database
Secured with the same API-key authentication as the REST endpoints
Available on all plans at no additional cost

MCP + Claude

# claude_desktop_config.json
{
  "mcpServers": {
    "reflexdb": {
      "url": "https://db_01J….reflexdb.cloud/mcp",
      "headers": {
        "Authorization": "Bearer rxk_live_…"
      }
    }
  }
}

# Claude now has live tools for your data
User:  "Top 5 products by revenue this month?"

[tool: query]
  query: orders(created_at > "2026-05-01") {
    product_id SUM(total_cents)
  } LIMIT 5

─── result (3 ms) ──────────────────
Widget Pro    $94,310
Basic Widget  $61,200

Enterprise

Built for regulated industries and large teams

Business and Enterprise plans include the controls that security, compliance, and procurement teams require.

Audit log export — full event history for every action
99.9% SLA (Teams), 99.95%+ SLA (Enterprise)
Dedicated support with 4-hour response (Enterprise)
Data Processing Agreement (DPA) included on all paid plans
Annual contracts with invoice billing (Enterprise)
Multi-user teams with seat management and invite flows
TOTP / MFA on all accounts

plan comparison

feature              Pro    Teams   Business    Ent.
────────────────────────────────────────────────────
Real-time CDC         ✓       ✓        ✓         ✓
MCP server            ✓       ✓        ✓         ✓
Multi-region          ✓       ✓        ✓         ✓
Custom subdomain              ✓        ✓         ✓
Audit log export                       ✓         ✓
SLA                 99.5%   99.9%    99.9%    99.95%+
Support             48 h    24 h      12 h      4 h

Ready to add a read layer to your stack?

Start with one free nano instance. No credit card required.

Start free View pricing

Everything your stackneeds from a read layer