Lizard
Get Started

Roadmap

Where the project actually stands — not a marketing timeline. See PLAN.md in the repo for the full phased design document this page summarizes.

Shipped today

  • Multi-engine connections — Postgres, MySQL, and MongoDB side by side, mixed freely.
  • Auto-generated, schema-derived CRUD UI with filters, search, and inline editing.
  • Virtual relationships across connections and engines, with no DDL on your databases.
  • Natural-language querying via an on-demand, tool-calling AI agent — schema exploration streams live, the generated SQL is always shown, and a run can be stopped mid-flight.
  • One-click and prompt-driven charts, plus dashboards that pin panels from different databases.
  • Dashboards as pages of blocks — table, chart, kanban, and gallery views over the same query, URL-backed variables (multi-select with an "All", and optional [[ ... ]] SQL blocks), an optional date range, tags with server-side search, save-as, JSON import/export, and either a drag-and-resize grid or a single block filling the whole page.
  • Type-aware filters, including whole-day datetime semantics (optional time pickers, half-open ranges that stay index-friendly).
  • Timezone-correct timestamps end to end — the browser's zone is validated server-side and applied as a session setting on Postgres, MySQL, and the federation engine.
  • Fleet-scale bounds throughout: scoped-only introspection, an enforced row limit on every query, bounded fan-out, a column-bounded catalog cache with stale-while-revalidate, idle connection-pool eviction, and paged, server-searched pickers.
  • CSV/JSON export honoring the current filters, sort, search, and column selection, with foreign keys optionally exported as a resolved "(label)" companion column. A result under the row-limit threshold streams straight back; anything larger becomes a background job you can leave running and cancel, tracked in Settings → Exports until you come back for it.
  • Role-based access (admin/editor/viewer), per-connection grants, and an audit log with tiered retention.
  • Credentials encrypted at rest (AES-256-GCM); all Lizard state lives in its own local SQLite file.

See Features for the full detail on each of these.

In progress — richer data console

Deepening the auto-generated console itself (still zero-config, still schema-derived — not adding a SQL/DBA client or AI features):

  • Rich type widgets — arrays, ranges, structured jsonb editing, bytea uploads, network types, PostGIS/pgvector-aware cells.
  • Inline cell editing, bulk edit/delete, row duplication, client-side constraint validation.
  • Saved views for tables themselves, plus calendar and self-referencing tree views (kanban and gallery already ship as dashboard blocks).
  • Many-to-many linked records auto-detected from junction tables.
  • Richer filter operators (regex, array contains/overlap, range overlap, jsonb containment).
  • Per-view auto-refresh, per-record comments, and record history from the audit log.

On the roadmap

  • SQLite as a fourth engine, behind the same driver abstraction as Postgres/MySQL/MongoDB.
  • Embeddings-based schema retrieval for very large fleets, so the AI agent scales past what fits in a name-only skeleton.
  • Saved AI "workflows" and scheduled reports.
  • A plugin system for custom widgets and dashboard panels.

Deliberately out of scope

Kept out on purpose, to stay a data console rather than something broader:

  • A low-code app builder — no drag-and-drop page composer.
  • A BI warehouse tool — no ETL, no materialized-view pipelines.
  • DBA/SQL-client tooling — no index/vacuum/activity/EXPLAIN browser, no function/trigger editor, no DDL against your databases.