I write code
After setup, supported coding-agent sessions can be captured and distilled into lessons for future work. Supported tools and recall behavior vary by integration.
Station online · capture running
Your AI work shouldn’t
start from zero.
Decisions, project history, and working context shouldn’t have to be reconstructed in every new session.
SpaceStation AI organizes work from supported AI tools in a Markdown vault on your computer, then brings relevant context into future work. Capture and recall behavior depends on the integration.
Claude Code · Codex CLI · Cursor · Grok Build · Claude Desktop via MCP
SCROLL TO EXPLORE
Two ways in
Keep useful context from your AI-assisted work.
I work with AI assistants
Keep useful decisions and project context from supported AI work in a local Markdown vault, ready to bring into future sessions.
Station core
The core every ship shares.
STATION_MAP // STANDBY
coremodulesfleetdocking bay
The fleet
SHIP-001 · docked2h ago
Atlas
2 decks · 15 subroutines · 1 rules
64 sessions · 8 meetings · 12 memories
SHIP-002 · docked5h ago
Beacon
1 decks · 6 subroutines · 0 rules
22 sessions · 3 meetings · 5 memories
SHIP-003 · docked3d ago
Corsair
3 decks · 21 subroutines · 2 rules
40 sessions · 11 meetings · 9 memories
Docking bay · unassigned
4 sessions · 2 meetings · 1 memories
Six verbs
The whole system, in six words.
Capture
After guided setup, supported integrations capture sessions automatically.
Distill
After sessions and during scheduled maintenance, the learning loop distills useful context into durable lessons.
Recall
Supported integrations can bring relevant past context into a conversation when it starts or when you request it.
Dock
Add a project to the station to organize its history and make it available to supported recall features. Meeting sync is optional.
Undock
Undock a project to stop bringing its knowledge into future work. Its files remain in your vault.
Debrief
Optional Granola and Wispr Flow connectors can sync meeting notes for project routing and debrief.
The graph
Everything you do connects.
Sessions, meetings and decisions can appear as nodes in a navigable graph, filterable by type or project. This illustrative view uses sample data; the graph view is available in the app.
Every note, session, and skill — one connected map.
KNOWLEDGE GRAPH
413 nodes
❯ filter nodes…⌘K
⚙ NODE CONTROLS
crm-portal
NET_MAP // 413 NODES · 619 LINKS
The learning loop
The station turns sessions into durable knowledge.
After sessions and during scheduled maintenance, a loop distills useful context into durable knowledge and logs each run in an Ops Log you can inspect.
- 01Review sessionsThe loop reviews captured sessions and identifies useful context.
- 02Distill lessonsIt writes durable, deduplicated lessons and updates a curated profile of how you actually work.
- 03Keep it where you can read itEvery run is logged to the Ops Log in your vault — plain Markdown, snapshotted to git if you want the history. Inspectable, diffable, yours.
Your vault
A folder before it is a product.
Station files are stored in a local Markdown vault that works with Obsidian. Connected AI tools and optional services handle the data you send to them. Back up your vault to protect against device or disk loss.
- Local filesStation files are stored in a Markdown vault on your machine. Connected AI tools and optional services handle the data you send to them.
- Plain MarkdownReadable today, readable in ten years. No lock-in, no database to escape.
- You decide what it knowsExclude a confidential meeting from recall, or trash a note and restore it later. Nothing is hard-deleted unless you confirm it.
- Control future contextUndock a project to stop including its knowledge in future work. Its files remain available in your vault.
Keep your work in files you can access.
Local Markdown files, compatible with Obsidian, and ready for you to back up.
Vault localCapture onUndock anytime