Observable

Executes JavaScript/TypeScript code in browser-based cells with automatic re-execution when upstream dependencies change, using a reactive dataflow graph to track variable references across cells. When a cell's inputs are modified, the runtime identifies all dependent cells and re-executes them in topological order, enabling live-updating visualizations and dashboards without manual refresh triggers.

Enables simultaneous code editing by multiple users on the same notebook with conflict resolution via operational transformation (or CRDT-based approach, not publicly documented). Changes from each editor are broadcast to all collaborators in real-time, with the platform handling merge conflicts when two users edit overlapping code regions. Version history is maintained at the notebook level, allowing rollback to any previous state.

Supports importing data from multiple sources: file uploads (format types not documented), cloud storage (specific services not documented), and web API endpoints. Data can be transformed using JavaScript/TypeScript in notebook cells, with support for common operations (filtering, grouping, aggregation) via standard JavaScript array methods or libraries like Lodash. Imported data is stored in notebook variables and can be visualized or queried reactively.

Maintains a complete version history of all notebook changes with commit metadata (author, timestamp, change summary). Users can view the history of any notebook, compare versions, and rollback to previous states. Version control is integrated at the notebook level (not cell-level), with automatic commits on save or manual commit creation. Available on all tiers (Free and Pro).

Organizes notebooks into workspaces with role-based access control (editor, viewer). Workspace owners can invite collaborators, assign roles, and manage guest access. Separate viewer tier ($10/month per viewer) allows read-only access to notebooks without editor permissions. Access control is enforced at the workspace level, with all notebooks in a workspace sharing the same access rules.

Connects directly to external databases (Snowflake, DuckDB, PostgreSQL, Databricks) from within notebook cells, executing SQL queries server-side and returning results to the browser for visualization. Connection credentials are stored securely on Observable servers, and query results are cached to avoid redundant database hits. Supports parameterized queries to enable interactive filtering without re-querying the entire dataset.

Provides an AI assistant (model identity unknown, likely GPT-4 or Claude) that generates JavaScript code for charts, data transformations, and analysis based on natural language prompts. The assistant has access to notebook context (previous cells, variable definitions, data schema) and can generate multi-cell workflows. Outputs are marked as 'inspectable' but the inspection mechanism is not documented — likely means generated code is visible and editable rather than a black box.

Provides built-in access to D3.js (open-source, 508M+ downloads) and Observable Plot (open-source charting library, 4.39M+ downloads) for creating interactive visualizations. D3 enables custom, low-level visualization control via SVG/Canvas manipulation, while Plot provides high-level declarative chart syntax for common chart types (bar, line, scatter, etc.). Both libraries are fully integrated into the notebook execution environment and can be combined with reactive cell dependencies for live-updating charts.

Publishes notebooks to shareable URLs with granular access control (public, private, or shared with specific users). Published notebooks can be embedded in external websites via iframe, exported as PNG images, or shared as standalone URLs. Observable watermark can be removed on Pro tier. Access control is managed at the notebook level, with separate viewer and editor roles. Published notebooks execute in the browser with the same reactive behavior as the authoring environment.

Executes notebooks on a schedule (daily, weekly, monthly, or custom cron expressions) and delivers results via email or webhook. Scheduled runs execute in the Observable cloud environment (not in user's browser), enabling long-running queries and large dataset processing without blocking the user. Results can be captured as PNG exports, JSON data, or full notebook snapshots. Pro tier only.

Provides a visual whiteboard environment (Canvas) for combining SQL queries, data tables, charts, freehand sketches, and notes in a single collaborative space. Canvases enable drag-and-drop composition of data analysis workflows without requiring code, supporting real-time collaboration similar to multiplayer notebooks. Specific technical capabilities (SQL support, data limits, visual query builder) are not documented.

Provides a library of pre-built UI components (buttons, sliders, dropdowns, text inputs, date pickers, tables) that integrate with the reactive notebook execution model. Components emit values that automatically trigger dependent cell re-execution, enabling interactive filtering and parameter adjustment without manual event handling. Components are open-source (4.39M+ downloads) and can be customized via CSS or JavaScript.

Open-source framework for building static data applications that can be deployed outside the Observable platform. Framework uses Observable notebooks as source files, compiles them to static HTML/JavaScript, and enables deployment to any web server or CDN. Supports the same D3/Plot visualization libraries and reactive cell model as Observable notebooks. Allows building data apps without vendor lock-in to the Observable platform.