Status
In private development. Early access by invitation.
Emolysis is the agentic lakehouse for data engineering and analytics. Direct a team of agents from your laptop to investigate business questions, develop dashboards, and build and test pipelines for your data platform. Underneath is a local-first coding agent that writes data pipelines and dashboards for the platform you already run, tests them on your machine, and deploys them as your platform's native units.
Use these docs to judge whether Emolysis fits how your team works. The status line at the top of each page shows where it stands.
The idea in one paragraph
Data engineering is more templatable than general software. About ten pipeline shapes cover most production work: ingesting files, appending incrementally, merging change feeds, keeping history, snapshots, backfills, late data, cleaning into silver, building gold metrics and pushing results back to applications. Each shape has a correct form on each platform. Emolysis starts from a tested template for the shape and platform, fills it with what it learns from your catalog and your request, and proves it with golden tests on DuckDB before it touches your platform.
The agent does not write pipelines from scratch. If a request does not match a shape, it says so instead of improvising code it cannot test.
What you work with
- Projects. A plain folder in your Git repository, with models, pipelines, tests and a project file that names the target platform. See Projects and emolysis.yml.
- Platform profiles. What the agent knows about Databricks, Snowflake or Microsoft Fabric: dialect, table format, merge idiom, deployment unit and known traps. See Platform profiles.
- Templates and golden tests. One template per shape and platform, each with fixtures and expected outputs. See Templates and slots and Golden tests on DuckDB.
- Dashboards. Charts built from a tested gold table, shared as a link when you choose. See Dashboards from a question.
- Surfaces. The desktop app, the
emocommand line, the MCP server for editors and coding agents, and checks on GitHub pull requests. They all run the same commands. See The emo CLI and The MCP server.
Where things stand
| Area | Status |
|---|---|
| Emolysis overall | In private development; no download, accounts or release date |
| Local engine on DuckDB | In development |
| Databricks | In development; the first platform |
| Snowflake, Microsoft Fabric | Planned, after Databricks; tested only on simulated services, not on a real account; no dates |
| PostgreSQL as a source | In development; read from, never deployed to |
| SQL Server as a source | Planned |
| Pipeline templates (ten shapes) | In development; none released |
Desktop app, emo CLI, MCP server | In development; not available |
| GitHub App | Planned |
Pull request checks in your own CI (emo check) | In development |
| Accounts, plans, shared dashboards | In development; nothing can be bought or shared yet |
| Template format and validator | Public source on GitHub, not packaged |
| Sample dashboard on this site | Live, on synthetic data; it renders the same chart specs the agent produces |
A note on your data
Local-first does not mean nothing leaves your machine. To write code, the agent sends prompts to the AI model provider you choose, and a prompt can include table and column names, schemas and sample values. With a paid plan's model usage, those prompts pass through Emolysis's metering service, which is designed not to store or log their content. A dashboard leaves your machine when you publish it, and only then. Read Data handling before you connect anything, and the security page for the summary.
How these docs are organized
- Start here. This page, How Emolysis works, Installation and the Quickstart.
- Build pipelines. Projects, pipeline shapes, templates and golden tests.
- Platforms. Profiles, connecting Databricks, Snowflake and Fabric, sources and deploying.
- Dashboards. From a question to a chart, sharing and chart types.
- Tools. The CLI, the MCP server, editors and coding agents, and GitHub checks.
- Account and data. Model keys and credits and data handling.
- Help. Troubleshooting and error codes.
Every page starts with a status line. Reference pages for commands, files and fields also carry a "Planned reference" label, because their names and formats can change before the first release.
Where to go next
Read How Emolysis works for the full loop, then the Quickstart for a first session. If you want the words first, the glossary defines every term these docs use, and the ten pipeline shapes each have their own page.