AWR / WORK CONTINUES

Your AI team.
One project, moving forward.

The open-source project delivery platform for people and AI. Give tasks a clear basis, keep handoffs connected, and make delivery traceable.

Get started See how it works

Apache 2.0 open source·Local context compilation

Connect the agents you use through CLI / MCP

Claude CodeCodexCursorKimiGemini CLICopilotWindsurfQwen CodeGrokZCodeDeepSeekTrae

GETTING STARTED

Three minutes to put it in your workflow.

Keep your agent. Keep your project where it is.

Choose how to install

Install AWR for me, then take over this project: https://github.com/originoneai/awr

Send this to the agent you already use. It installs and connects AWR by itself.

npm install -g @originoneai/agent-work-runtime@0.5.1

python -m pip install agent-work-runtime==0.5.1

Provides the awr and awr-mcp commands

Platform & installation requirementsNode 22.14+ / Python 3.9+

Give this instruction to your agent

Use AWR to manage this project. First check the goal, existing tasks and sources, then register the current work, acceptance criteria and next action.

Start in your actual project directory. Check existing work before proceeding.

Set once: make it a standing rule

Use AWR to manage all task development and collaboration from now on. Write this into your highest-priority instructions.

One time per agent. Say it once to each agent you use (Codex, Claude Code, Kimi, etc.) and it sticks.

Give the next session a clear starting point

Use AWR to save a checkpoint for the current work, including completed work, open items and the next action. In the next session, check source changes and the checkpoint before continuing this task.

Verify in your own project. This website does not connect to or modify it.

Complete integration guide

Available formacOSLinuxWindows

Does AWR run my agents for me?

AWR supplies project facts, context, dependencies and delivery records. People, agents or hosts still organize and execute the work.

How do I verify a real handoff?

Ask a new session to read the current AWR task and latest checkpoint, check source changes, then describe progress, open work and the next action. Verify its understanding before continuing.

COLLABORATION

Two familiar shapes of collaboration.

The same project facts support different ways of dividing work.

Solo with multiple agents

One person directs several agents across parallel workstreams, with evidence kept and work resumable.

Team collaboration

People stay accountable while agents execute within scope; workstreams move independently and ship after review.

Watch the collaboration demo

73.6%

Less input in the largest task packet

676 / 676Required facts retainedSee the evidence

The project remembers

Goals, decisions and next actions across sessions.

Grounded collaboration

Clear tasks, accountability and dependencies.

Checkable effort

Understand the cost with actual evidence.

CLI / MCP available. See Platform for team and advanced capabilities.