Image generation
Generate images from prompts inside the platform, with model selection and per-generation settings. Outputs land directly in your gallery.
Platform
A transparent view of the platform: what is live, what is in active development, and what is planned next. Updated as work ships.
Live on the platform today.
Generate images from prompts inside the platform, with model selection and per-generation settings. Outputs land directly in your gallery.
Text-to-speech with multiple voices and style direction. Produce narration and audio clips without leaving your workspace.
Every generated asset is saved, previewable, and organized into folders. Move, rename, and batch-manage outputs across projects.
Persistent memory your agents read and write across sessions. Preferences, project context, and prior decisions carry forward automatically.
Full-text search across assets, conversations, and memory. Find any generation or note from a single search bar.
A Model Context Protocol server that exposes the platform to any MCP client. Drive canvases, assets, and generation from Claude Code or your own agents.
First-party connectors that let agents pull from and push to external services, with scoped permissions per connection.
Run history, token usage, latency, and per-step traces for every agent action. See exactly what ran, what it cost, and where it failed.
In active development.
Prompt-to-video alongside the existing image and voice pipelines. Generated clips save to the gallery with the same folder and search support.
Describe an app and watch it built and executed in an isolated Vercel Sandbox. Iterate on running code with live previews before deploying.
An infinite canvas that agents control over MCP. Documents, images, and video stream onto the board in real time as agents work.
Share a canvas, asset, or app with a public link. Read-only by default, with revocable URLs and no account required to view.
On the near-term horizon.
An agent-maintained wiki built from your memory bank and assets. Structured, linked pages that stay current as your projects evolve.
Shared workspaces with roles and permissions. Common galleries, shared memory, and team-level usage reporting.
Expanding the connector catalog based on demand: issue trackers, docs platforms, cloud storage, and analytics sources.
Build and register your own connectors with a typed SDK, so agents can reach internal tools and private APIs.
Dates are targets, not commitments. Priorities shift based on what builders actually use, so shipped work is the most reliable signal of direction.