Fivetran, Airbyte, Stitch, and HevoData work from pre-built connector catalogs — if the source you need isn't in the catalog, you wait. rsync.ai generates the connector on demand from any API's docs, billed on data volume, not per row or Monthly Active Row — with source-available connector code to read or fork.
Instead of waiting for a vendor to add a source to a catalog, rsync.ai generates the connector from the API's docs in minutes — with source-available code to read, audit, or fork.
Point rsync.ai at any REST or GraphQL API documentation and it produces a working connector in minutes. No waiting for a vendor to add it to a catalog.
Every generated connector is readable code under the rsync.ai Source-Available License — read it, audit it, or modify it when you self-host. Fivetran's connectors are closed; you never see the code.
The pipeline discovers the schema, detects PII, plans the sync, and asks for your approval before it runs — no YAML to write, no connector UI to wire up.
Including where competitors are still ahead: catalog breadth, community size, and maturity.
| Feature | rsync.aiGenerated connectors | Fivetran | Airbyte | Stitch | HevoData |
|---|---|---|---|---|---|
Connector model How connectors are built and what drives them | Generated on demand from docs | Pre-built catalog (700+) | Community catalog (600+) | Pre-built catalog (140+) | Pre-built catalog (150+) |
Plain-English pipeline setup | |||||
Generate a connector from any docs URL Point at REST or GraphQL API docs and get a working connector in minutes, on rsync.ai Cloud; Fivetran's AI Connector Agent is in private preview and covers REST APIs only | partial | partial | |||
Plain-English questions that write SQL Ask rsync.ai writes the SQL from a question; you review it and run it in the Explorer | |||||
Real-time log-based CDC | partial | ||||
Self-hostable / on-premise rsync.ai self-hosting is available now (source-available, rsync.ai Source-Available License); Fivetran Hybrid Deployment keeps processing in your env but isn't full self-host | partial | ||||
Source-available connector code Can you read and audit the connector code? rsync.ai: yes, under its Source-Available License. Stitch's Singer taps are public on GitHub. | partial | ||||
No per-row / per-MAR pricing Airbyte self-host is free; Airbyte Cloud bills on data volume (per GB / per million rows) | partial | ||||
You approve the plan before any pipeline runs No pipeline is created or started until you confirm the plan | |||||
PII detection before write Scans for likely personal data, then mask, hash, or drop fields before data lands in the destination | partial | partial | partial | ||
AI / LLM API included rsync.ai includes OpenAI on all managed plans — no extra key needed | |||||
One-command Docker self-host Ships with rsync.ai self-hosting (Kubernetes Helm chart in preview); Airbyte deprecated Docker Compose — self-managed Airbyte now runs on Kubernetes via abctl (a local kind cluster) or Helm |
How a tool meters usage matters as much as the sticker price. Per-row and per-MAR models bill on rows; rsync.ai bills on GB moved.
rsync.aiyou
Per GB moved (10 GB free)
Billed on GB moved, not per row or MAR. Each plan includes a monthly GB allowance; pipelines pause at your cap.
Fivetran
Per Monthly Active Row (MAR)
Counts each distinct row you add, change or delete once a month; the initial historical sync is free. Bills follow how many distinct rows change.
Airbyte
Volume-based credits (Cloud) / Free (self-host)
Cloud bills on data volume (per GB / per million rows), so backfills still add cost. Self-host is free but you own all ops.
Stitch
Per row synced
Billed on rows replicated per month, in tiers.
HevoData
Per event
Billed on events ingested per month, in tiers.
Want a deeper look at a specific tool?
rsync.ai generates a connector on demand from an API's documentation — it discovers the protocol, auth, schema, and operations, and produces a working connector in minutes; the connector code is source-available to read, audit, or fork. Fivetran and Airbyte connectors are catalog items: pre-built, statically configured, and maintained by a vendor. If the source you need isn't in the catalog, you wait. With rsync.ai, you point at the docs and get one in minutes. (Under the hood the connectors implement MCP, the open standard for AI tool use, so the pipeline can run them natively — but you don't need to think about the protocol to use them.)
For the core use cases — replicating databases and SaaS sources into a warehouse, lake, or another database — yes. rsync.ai covers PostgreSQL, MySQL, SQL Server, and MongoDB with real-time CDC, plus Shopify, Google Sheets, Amazon S3, Google Cloud Storage, Azure Blob Storage, and the BigQuery and Databricks warehouses — all live. Snowflake, Redshift, Stripe, and GitHub are in preview. Where Fivetran is still ahead is catalog breadth: 700+ managed connectors. rsync.ai closes that gap with on-demand connector generation on rsync.ai Cloud — you're not blocked waiting for a vendor to add a source.
Yes — rsync.ai's self-hosting is available now: a single install command brings up the full stack (pipelines, CDC, AI services, SQL Explorer) via Docker Compose in your own VPC; a Kubernetes Helm chart is in preview. Airbyte deprecated Docker Compose; self-managed Airbyte now runs on Kubernetes via abctl (a local kind cluster) or Helm, with multiple coordinated services — server, workers, scheduler, its own database. Both are source-available: Airbyte's platform is under the Elastic License 2.0 (its protocol is MIT), and rsync.ai is under the rsync.ai Source-Available License — free to self-host and modify for your company's own internal use, with a commercial license needed to resell it or host it for others. AI connector generation runs on rsync.ai Cloud, which is also live and free to start.
Fivetran bills by Monthly Active Rows: each distinct row you add, change, or delete counts once a month, and the initial historical sync is free. Stitch bills by rows replicated. rsync.ai's hosted plans bill on GB moved, so your bill tracks bytes transferred, not row count. Each plan includes a monthly GB allowance (10 GB on the free month), and pipelines pause at your cap.
Hosted managed plans include OpenAI — no key needed. On self-hosted deployments, you bring your own key (OpenAI) or run Ollama locally at $0. Self-hosted, the AI layer powers plain-English pipeline setup, and Ask rsync.ai once you switch it on and give it a Claude, ChatGPT or Gemini key; AI connector generation runs on rsync.ai Cloud. Without it, the core pipeline engine (CDC, batch sync, schema discovery) still runs.
One month free. On-demand connector generation on every plan. No per-row pricing. Cancel any time.