Reads user intent, context, files, goals, and signals.
~/zephora/welcome.sh
Stop repeating prompts.
Build with an agent that executes.
Zephora AI is an autonomous command layer that understands your intent, plans the workflow, runs actions with tools, verifies outcomes, and returns clear receipts.
01
~/zephora/problem.md
Assistants reply. Zephora executes.
Most AI tools stop at text. You still copy, paste, check, deploy, and repeat — every session starts from zero. Zephora becomes a persistent execution layer that remembers the goal and carries it to a verified outcome.
- Prompting resets every session
+ Zephora keeps context, rules, and goals persistent
- AI gives drafts
+ Zephora produces verified action plans and receipts
- Users manually connect every step
+ Zephora connects tasks, tools, APIs, and workflows
- There is no record of decisions
+ Every action has a traceable log
// assistants reply. zephora executes.
02
~/zephora/agent-loop/
The agent loop
A closed loop of executable steps. Intent goes in, a verified outcome and receipt come out.
Breaks the goal into steps, dependencies, tools, and risk checks.
Runs the workflow through connected APIs, bots, scripts, or internal tools.
Checks whether the output matches the original goal.
Creates a clear action receipt with status, logs, and next steps.
// agents do tasks. zephora manages outcomes.
03
~/zephora/receipt.log
Every run ends with a receipt
receipt #2026-001
status: verified
- intent
- "Research new market signals"
- context
- "User requested early trend discovery"
- plan
- "Collect sources → filter signals → summarize opportunities"
- action
- "Research completed"
- verification
- "Sources checked, summary generated"
- approval
- "User review required before publishing"
- rollback
- "Available"
04
~/zephora/policy.cfg
Authority modes
You set how much rope the agent gets. Dial it up as trust builds.
05
~/zephora/memory/
Memory & ownership
Zephora keeps portable, versioned, source-tracked memory. It belongs to you.
⇄portable
User context can move across models and tools.
⌛versioned
Every goal, rule, and memory update has a timestamp.
⌖sourced
Important facts are linked to where they came from.
✎editable
Users can correct, remove, or update memory anytime.
06
~/zephora/use-cases.md
One layer, many agents
Builder Agent
Generate product plans, repo tasks, API docs, and launch checklists.
Research Agent
Track signals, summarize sources, compare projects, and produce reports.
Crypto Agent
Scan tokens, monitor launches, analyze contracts, and create safety notes.
Creator Agent
Turn ideas into posts, threads, scripts, images, and launch content.
Team Operator
Automate repetitive workflows, updates, reminders, and task routing.
Personal OS
Maintain goals, rules, memory, and long-running projects.
07
~/zephora/CHANGELOG
Roadmap
✓v0.1Terminal Interfaceshipped
✓v0.2Agent Planning Engineshipped
✓v0.3Memory Layershipped
●v0.4Tool Executionin progress
○v0.5API + Bot Integrationsplanned
○v0.6Autonomous Workflowsplanned
○v1.0Zephora Agent OSgoal
08
~/zephora/playground/
Run it yourself
Type a command — or tap one — and watch Zephora turn intent into a receipt.
guest@zephora:~/playground$ help
try: build landing page · scan token · summarize docs · create workflow · research trend · generate launch plan
09
~/zephora/access.sh
Request early access
Join the early operators building with Zephora AI.
$ ./access.sh --request