Agnostic streamlines the entire data lifecycle with a set of open-source tools—available on our Platform or for self-hosted deployment:
- AGT — a SQL-driven ETL engine that turns ClickHouse into a transformation runtime.
- AGP — an asynchronous query processor that schedules, runs, and serves heavy SQL workloads on ClickHouse.
- AGX — a modern desktop/web interface for exploring, querying, and visualizing data.
Together, they give teams one coherent path from raw sources to insights — self-hosted or fully managed in the cloud.
AGT
SQL-Driven ETL on ClickHouse
What it is. AGT makes ClickHouse act as a high-performance, SQL-native transformation engine for batch and micro-batch pipelines. All logic lives in SQL; no separate DSL or orchestrator is required.
How it works.
- Sources are SQL queries; each returned row becomes a task flowing through the pipeline.
- Variables (
vars) are key-value pairs you can inject and update across stages. - Processors implement transformations, including
apply(run SQL per task), orderedsequence, andaccumulate(buffer then finalize a batch). - Benefits: fast/columnar execution, deterministic runs, and embeddable/standalone deployment.
AGP
Asynchronous Query Processor
What it is. A service that manages long-running analytical SQL against a ClickHouse cluster using a queue + workers model, with status tracking and persisted results. It uses PostgreSQL for orchestration and stores successful results in an object store or local FS.
Key capabilities.
- Executions progress through PENDING/RUNNING/CANCELED/FAILED/SUCCEEDED with progress and result availability.
- Execution collapsing: deduplicate identical queries using a
query_id. - Tier-based resource allocation: assign tiers to control CPU/time and route to appropriate worker pools.
- Components: API server (create/list/poll/retrieve), Workers (run queries), Bookkeeper (recover dead workers, expire old executions).
- API docs: embedded Swagger UI at
/v1/async/docs.
AGX
Explore & Query
What it is. A desktop app (and web UI) to explore and query data, built with Tauri, SvelteKit, and Plot. It can run locally (native) or connect to a remote server.
Features.
- Interactive SQL editor (syntax highlighting) with LLM integration and a schema browser.
- Tabular result viewer, drag-and-drop files, and cross-platform support (macOS, Linux, Windows).
- Live mode at
agx.app, Docker compose for local runs, optional Ollama integration, and a one-liner to install Agnostic ClickHouse UDFs.
The Agnostic Platform (Managed)
Prefer not to run anything yourself? Agnostic provides a managed offering: spin up cloud/data pipelines on hosted infrastructure with automatic scaling (ClickHouse + Apache Iceberg) and attach compute instances sized to your needs. This brings the suite together without DevOps overhead. Agnostic
How They Fit
| Layer | Role | You Get |
|---|---|---|
| AGT | SQL-native ETL/transforms on ClickHouse | Declarative pipelines, batch/micro-batch, deterministic runs |
| AGP | Orchestration of heavy/long SQL | Queued async execution, dedup, persisted results, tiering |
| AGX | Human interface | Querying, schema discovery, visualization, AI assist |
| Managed Platform | Hosted runtime | Pipelines + compute on managed infra, autoscaling, no servers to maintain |
This stack lets you compose reliable pipelines (AGT), schedule and serve heavyweight analytics (AGP), and put insights in users’ hands (AGX) — self-hosted or fully managed.