Astro is a self-hosted application layer for AI. It gives you structured memory (markdown, documents, tables, diagrams), automation (Python scripts and schedules, Slack agent tasks), and a full MCP + REST surface so models and humans operate on the same data—not parallel silos. Universes split projects cleanly; each one gets its own dashboard, categories, and agent context.
Memory, automation, and coordination in one Docker install—so you spend time on product logic, not gluing vector DBs, cron runners, and Slack bots together.
Markdowns and documents embed into local ChromaDB with CPU-based vectors—no cloud embedding keys. Semantic search spans the full corpus so retrieval stays private, fast, and grounded in your content.
Rich notes with formatting, images, categories, and full-text search. Each note feeds retrieval automatically so agents draw from the same material you see in the UI.
Upload PDF, DOCX, XLSX, PPTX, TXT, MD, or CSV; Astro ingests and embeds each file. Uploads become searchable memory instead of passive attachments on disk.
Create and edit diagrams with
Excalidraw:
shapes, arrows, and sketches in a full embedded editor.
Import and export .excalidraw files and keep visuals next to your specs.
Spreadsheet-style grids with typed columns, inline row editing, and CSV import or export. Model structured datasets that both people and automation can read and update reliably.
Save URLs with titles and categories beside notes and documents. Links stay discoverable in search so references surface without digging through browser history.
Connect a Slack bot so Agent Tasks post instructions to your workspace channels. Coordinate people and AI agents in Slack while Astro handles scheduling and delivery.
Bind a task to a markdown note and Slack channel: run now, on a cron schedule, or once later. Astro sends instructions through a configurable template with API context and note IDs.
Each universe opens to a four-column markdown widget board: status, links, images, runbooks. Drag to reorder; agents upsert widgets by tag over MCP. Live refresh keeps the board current.
Python as first-class content: syntax-highlighted editor, autosave, run on the server with captured output. Call APIs, scrape pages, or glue Astro data without leaving the app.
Schedule scripts like cron jobs: manual run, recurring UTC cron, or one-shot future time. Last status and stdout/stderr stay on the task—automation you can inspect.
A built-in Model Context Protocol server exposes tools for search, notes, tables, diagrams, scripts, Python tasks, dashboard widgets, Slack agent tasks, and more. Point any MCP-compatible client at your instance so agents read and write the same corpus you see in the UI.
Watch videos to see what Astro can do.
Astro separates coordination from execution. Agent Tasks push markdown instructions to Slack so people and external agents pick up work in the channel where decisions happen. Python Tasks run saved scripts on the server—scraping, API calls, transforms—on a schedule you control. The universe dashboard shows what matters now. Same database, same MCP tools, different rhythms.
Categories, pins, and hierarchical folders keep the corpus navigable for humans while agents use search and IDs. You are not filing notes for AI—you are building a substrate AI apps can assume.
Browse DocsParent and child tree across content types.
Markdowns, scripts, documents, diagrams, tables, and links one click away.
Scheduled prompts to Slack from markdown notes.
Cron and one-shot runs of server-side scripts.
Per-universe markdown boards agents can update by tag.
An AI application stack: memory, automation, Slack coordination, and MCP—in one self-hosted install.
One workspace. Many universes. Build on top of AI, not around it.
