Agent0 is an AI agent framework built on Obsidian and Claude Code. It gives you 7 specialized AI agents powered by 83 reusable skills, with built-in governance, persistent memory, and a vault that catches every idea and ships the ones that matter.

Do I need to know how to code to use Agent0?

No. The Starter tier uses copy-paste prompts with step-by-step instructions. The Pro tier runs one command. The Custom tier is fully set up for you.

How is Agent0 different from other AI agent frameworks?

Agent0 runs on Obsidian and Claude Code with your data staying local. No hosted servers, no monthly infrastructure costs. The reflect/evolve layer (seeing your thinking patterns over time) is something no other framework offers.

Scheduled Agents

Cortex and Radar work while you sleep.

How It Works

Two agents run unattended via the RemoteTrigger API, up to 3 runs per day. Radar handles outbound work: monitoring repositories, detecting signals, flagging captures. Cortex handles inbound work: proposing tasks, writing journal entries, running reflections.

There is no separate "heartbeat" process. It is just Cortex and Radar working on a schedule, each with clearly defined roles and boundaries.

The Schedule

Role	Agent	When	What It Does
Scout	Radar	Every run	Monitor repos, detect market signals, flag captures
Task Proposals	Cortex	Every run	Propose up to 5 tasks based on vault state
Chronicle	Cortex	10pm ET	Write nightly journal entry, update memory
Mirror	Cortex	Sunday midnight	Weekly reflection on patterns and progress
Compass	Cortex	1st of month midnight	Monthly reflection on direction and goals

Roles

Each role is a focused job that runs within a scheduled agent session.

Scout: Radar scans monitored GitHub repos for new releases, breaking changes, and signals worth capturing. Results go to knowledge-base/repos/ and radar/.
Task Proposals: Cortex reviews the vault state and proposes up to 5 actionable tasks. All proposals land in tasks/backlog/ and wait for human approval.
Chronicle: Cortex writes a nightly journal entry summarizing what happened that day. Updates memory with key learnings.
Mirror: Cortex writes a weekly reflection. Surfaces patterns, recurring themes, and progress against goals.
Compass: Cortex writes a monthly reflection. Evaluates direction, checks alignment with long-term goals, and suggests course corrections.

What They Can Touch

Scheduled agents have strict boundaries. They can only write to safe zones.

Category	Paths
Allowed	daily/, tasks/backlog/, tasks/Task Board.md (Backlog column only), .claude/memory/, knowledge-base/repos/, radar/
Protected	CLAUDE.md, .claude/skills/, governance/, projects/

Auto-Merge Rules

When you run /sync, it checks open agent PRs and auto-merges the safe ones. A PR qualifies for auto-merge when it meets two conditions.

PR title starts with Cortex: or Agent:
All changed files are within the allowed safe zones listed above

If a PR touches any protected file, it requires manual review. The system never auto-merges changes to skills, governance, or project files.

PreviousMemory NextHooks