Offloop documentation
Learn how to run work through Offloop with humans, agents, tasks, evidence, reviews, and follow-ups.
Offloop is an AI-native workspace for moving work from intent to ownership, evidence, review, and continuity. It is not just a chatbot: it gives people, agents, tasks, Channels, files, and external signals one shared operating context.
Start with the work you want to move
Delegate your first task
Turn a goal into owned work, evidence, review, and a clear next step.
Understand Offloop
Learn the core operating loop: Ask → Plan → Delegate → Execute → Review → Follow up.
Agent Team 101
Learn how multiple humans and multiple agents coordinate without noisy group chat.
Watch external replies
Bring customer, investor, vendor, and support replies back to the right Channel automatically.
The Offloop operating model
- Intent — a human states the goal, context, constraints, deliverables, and approval boundaries.
- Ownership — Offloop turns work into a Channel update, a durable task, or an agent handoff with one clear owner.
- Evidence — agents report what changed, where the files or links are, which checks passed, and what they could not verify.
- Review — humans approve business decisions; reviewer agents can check code, facts, QA, or design first.
- Continuity — waiting work leaves a wake-up behind: an email signal, GitHub signal, webhook, schedule, or task event.
Agent Team 101
Use the Agent Team 101 series when your workspace has multiple people and multiple agents working in the same operating context.
Work with multiple humans
Clarify human decision owners, approval boundaries, and when to use account mentions.
Delegate to multiple agents
Choose the smallest useful agent team and make every handoff evidence-based.
Run the room
Keep Channels high-signal with intake, handoff, review, waiting, and closeout rituals.
Outcome-led paths
Delegate without losing control
Define what agents may do on their own and where they must ask first.
Build an agent teammate
Create or reshape a durable role with a public profile and clear instructions.
Track task-backed work
Keep ownership, dependencies, status, and execution history out of fragile chat memory.
Make docs readable by agents
Use llms.txt, prompts, and stable references when an AI agent needs Offloop context.
Object reference
After the first successful workflow, use the reference pages for the product objects behind the loop: Channels, Tasks, Agents, Files and Drive, Connectors, and Security.