# for-ai-agents
Task management
for AI agents.
AI drives it from the CLI; you watch over it in the GUI. Even a freshly started AI catches up on the context by itself — from the records.
// for humans — GUI
You survey it — board, timeline, dependency graph. The AI drives; you watch over it.
# the-problem
An AI's memory is a black box. You can't see what it remembers. Switch sessions, and the AI has forgotten yesterday's “why.”
→ Because the context is trapped in the AI's black-box memory.
# how-it-works
Put the context in records you can read too.
Tasks, decision records, dependencies. The AI reads them and reaches the context it needs on its own.
// context
The AI catches up by itself
It searches and filters tasks (what), decisions (why), and dependencies (how they connect). Even a fresh AI — no need for you to re-explain.
// yours
Stays on your machine
Records live on your own device. Nothing is sent over the network; it works offline. Export to JSON / Markdown / CSV anytime.
// transparent
amenbo doesn't write
The AI and you do the writing. amenbo itself never rewrites or adds to your data. Who wrote what (you / AI) is kept, too.
# two-pillars
Tasks, and decision records.
Tasks are what to build next. Decision records are why. amenbo keeps both as first-class records — linked by dependencies, found by filtering.
Tasks — what, in what order
Decisions — why it was done
// dependencies
Tasks and decisions connect by dependency edges. Trace them with blocked by and blocks; finish one and the next comes alive.
// find
--filter by status, priority, assignee, or dependency. Pull just the records you need out of a bloated history.
# your-data
Your data is yours.
Having no server is, in itself, the reassurance.
- Local-only. Data never leaves your device; nothing is sent over the network.
- The AI and you do the writing. amenbo itself never touches your data.
- Export to JSON / Markdown / CSV anytime.
- Nothing lost, nothing broken, always recoverable.