Documentation

rsync.ai is a plain-English data pipeline. You describe what you want in plain English, and the agent handles the rest — generating missing MCP connectors, moving data, and letting you query results — all in one product.

Quick start

1. 1Sign up free at app.rsync.ai — nothing to provision, start building immediately. (Self-hosting on Docker Compose arrives July 2026.)
2. 2Open the pipeline builder and connect your first source.
3. 3Describe your sync: "Sync Shopify orders to Postgres every hour."
4. 4If the MCP connector doesn't exist, the Tool Generator builds one from the source's API docs in 2–5 minutes.
5. 5Approve the MCP connector schema and start the sync.
6. 6Open the Data Explorer to query your synced data in plain English.

A pipeline in rsync.ai moves data from a source to a destination on a schedule or in real time. You configure it in plain English — no YAML, no JSON, no UI forms.

How a pipeline runs

Supported sources

Destinations

• PostgreSQL (self-managed or cloud)
• MySQL
• Google Sheets
• AWS S3, GCS, Azure Blob, MinIO — byte-identical blob passthrough for any file format
• Snowflake, BigQuery, Redshift — coming soon

Schedule options

• Real-time (CDC) — Postgres & MySQL only
• Cron schedule — e.g. every 15 minutes, daily at 6am UTC
• Manual trigger — run on demand from the UI or API

The Tool Generator is rsync.ai's AI agent that builds a production-ready MCP connector from any REST or GraphQL docs URL. No code to write — paste the docs link and the agent produces a versioned, containerized MCP connector in minutes.

What gets generated

• Authentication handler — API key, OAuth 2.0, bearer token, basic auth
• Schema discovery — introspects all available resources and fields
• Cursor pagination — offset, page-token, and link-header styles
• Rate-limit backoff with configurable retry budget
• Dockerfile for isolated, reproducible execution
• Versioned output — regenerate at any time if the API changes

How to generate an MCP connector

1. 1In the pipeline builder, click "Add source" and paste the API docs URL (e.g. https://developers.notion.com/reference).
2. 2rsync.ai fetches the spec, identifies endpoints, and shows a preview of the resources it will sync.
3. 3Review and approve the MCP connector schema — your human-in-the-loop gate before any data moves.
4. 4The MCP connector generates and is immediately available. Typical time: 2–5 minutes.

After a pipeline runs, query the synced data directly in rsync.ai — no BI tool, no separate database client. Type a question in plain English and the Explorer converts it to SQL and runs it instantly.

Example session

select title, days_open
from linear_issues
order by days_open desc
limit 10;

Features

• NL→SQL: describe your question, Explorer writes the query
• CodeMirror 6 editor — syntax highlighting, autocomplete, keyboard shortcuts
• Runs against the destination database that received your synced data — no copy, no extra latency
• Export as CSV or JSON
• Query history — re-run or iterate on past queries

Everything you can do in the rsync.ai UI — create pipelines, trigger runs, check status, and query results — is available over a REST API, so you can wire pipelines into CI, orchestrators, or your own apps.

Authentication

Create a token in app.rsync.ai → Settings → API tokens and send it as a bearer token. The base URL is https://api.rsync.ai/v1.

# Trigger a run for an existing pipeline
curl -X POST https://api.rsync.ai/v1/pipelines/pl_123/runs \
  -H "Authorization: Bearer $RSYNC_API_TOKEN" \
  -H "Content-Type: application/json"

# Poll the run status
curl https://api.rsync.ai/v1/runs/run_456 \
  -H "Authorization: Bearer $RSYNC_API_TOKEN"

Core resources

Change Data Capture (CDC) streams row-level inserts, updates, and deletes from your database in real time using a Debezium-based log reader. Supported for Postgres and MySQL. Other sources use scheduled batch sync.

Postgres setup

-- Enable logical replication (requires superuser)
ALTER SYSTEM SET wal_level = logical;
SELECT pg_reload_conf();

-- Create a replication slot
SELECT pg_create_logical_replication_slot('rsync_slot', 'pgoutput');

-- Grant permissions to the rsync user
CREATE ROLE rsync_replicator WITH REPLICATION LOGIN PASSWORD 'your-password';
GRANT SELECT ON ALL TABLES IN SCHEMA public TO rsync_replicator;

MySQL setup

-- Add to my.cnf / my.ini
[mysqld]
log_bin       = mysql-bin
binlog_format = ROW
binlog_row_image = FULL
server-id     = 1

-- Create replication user
CREATE USER 'rsync'@'%' IDENTIFIED BY 'your-password';
GRANT SELECT, REPLICATION SLAVE, REPLICATION CLIENT ON *.* TO 'rsync'@'%';

rsync.ai can move any file format between cloud object stores — byte-identical, no parsing required. PDFs, images, Parquet files, video, archives, or any binary — if it lives in S3, GCS, or Azure Blob, rsync.ai can copy it anywhere else in the same class.

Supported pairs (v1)

Metadata catalog

Alongside every file transfer, rsync.ai writes a metadata record to your destination so you can track what was moved and detect changes for incremental re-sync:

Limits

• Maximum file size: 1.5 GB per object
• No limit on number of objects per pipeline run
• SHA-256 integrity verified on every transfer — mismatches fail loudly

Coming soon

• ◦Parse-for-AI — extract structured text from PDFs and images, chunk, and send to a vector store (separate lane from blob passthrough)
• ◦Warehouse blob destinations — land files into Snowflake internal stages, BigQuery Cloud Storage, Redshift Spectrum (requires warehouse feature)

When it ships, rsync.ai will be source-available under ELv2 — free to run on your own infrastructure. The stack is Docker Compose with three core services: the pipeline agent, the API server, and a Postgres metadata store.

Requirements

• Docker Engine 24+ and Docker Compose v2
• 6 GB RAM minimum (8 GB recommended)
• 20 GB disk space
• OpenAI API key or Ollama running locally

Environment

# .env  (copy from .env.example)
OPENAI_API_KEY=sk-...           # or set OLLAMA_BASE_URL instead
POSTGRES_PASSWORD=changeme       # internal metadata store — change in prod
SECRET_KEY=change-in-production  # session signing key

# Optional
S3_BUCKET=rsync-data             # for pipeline run artifacts
S3_ENDPOINT=https://...          # MinIO, Cloudflare R2, etc.
TRUST_PROXY=1                    # set if behind nginx / Caddy for TLS

Start up

git clone https://github.com/rsync-ai/rsync
cd rsync
cp .env.example .env    # fill in your values
docker compose up -d

The UI is at http://localhost:8080. Put Caddy or nginx in front for TLS in production.

rsync.ai is source-available under the Elastic License 2.0 (ELv2). You can read, run, and modify the code for your own internal use. You may not offer rsync.ai as a managed service to third parties without a commercial agreement.

When self-hosting ships (July 2026) you'll be able to run it freely for internal use. For OEM or multi-tenant deployments, contact us.