lizhenjie137/intelligent-organization
ProductivityA local-first workspace where people and AI agents split tasks by declared capabilities, with explicit handoffs and acceptance.
Intelligent Organization
Intelligent Organization is a local workspace for human–AI collaboration. It turns a shared goal into tasks that have an owner, inputs and outputs, and acceptance criteria, so people contribute where they are strongest and agents handle analysis, drafting and organizing. The product's long-term vision is multi-person collaboration without a fixed office; the current version only closes the loop inside one local workspace.
What it does
- Shared goals: each task keeps its background, expected result and completion criteria.
- Clear ownership: tasks are assigned to specific people or agents from a full roster; missing capabilities, permissions or model conditions are shown rather than hidden.
- Visible process: claiming, starting, publishing progress, submitting, accepting and returning are recorded separately.
- Traceable handoffs: finished upstream results with evidence and next steps are passed downstream, and the receiver can accept or return them.
- Capability catalog: people declare skills with source evidence and admin verification; agents register declared capabilities. A skill file is a structured description (name, inputs, outputs, acceptance, steps, boundaries, tools, examples) imported from Markdown/
SKILL.mdfrontmatter or a defined JSON structure.
Model access is configured by an admin as either a local model service or a compatible cloud API. Agents in the current version work on supplied text only; they do not browse the web, run scripts, send messages or call external business systems.
Quick start
The desktop client is called 小辣椒 (Xiaolajiao). For the installer route, download the Chinese or English Windows x64 installer from the v0.1.0 release, run it, and register with a display name, username and a password of at least 12 characters. No email is required. The first account in an empty workspace becomes the admin. Then create a goal, add member capabilities or materials, assign tasks, and configure a model if agents are needed.
For the source route, download the language-specific source ZIP or clone the repository, enter the language edition directory, then run in PowerShell:
.\Install.ps1
.\Start-Xiaolajiao.ps1
The application opens in the browser at http://127.0.0.1:8769 for Chinese or http://127.0.0.1:8771 for English. If a port is occupied, the start script stops; pass a free port instead:
.\Start-Xiaolajiao.ps1 -Port 8782
Source runs need Windows, Python 3.11+ and PowerShell; building the desktop client additionally needs Node.js 22.12+. Source data lives under data/workspace/ in the package; the installer stores data under %LOCALAPPDATA%\Xiaolajiao\zh\workspace or en\workspace. Installer and source editions, and Chinese and English editions, keep separate data and do not migrate automatically.
Constraints
- Only Windows x64 is supported today; the service listens on local
127.0.0.1, so cross-device, LAN and internet collaboration are not available. - Multi-account collaboration works on the same computer and workspace only.
- Cloud model use sends task materials to the configured provider and may incur provider fees; the agent budget counts invocation attempts, including failures, not tokens or money.
- Uploading a skill does not install tools, run scripts or grant permissions. People cannot accept tasks assigned to others, and submitters cannot accept their own human deliverables.
- Imported documents are snapshots and do not track later file changes.
- The license is source-available: download, install, use and secondary development are allowed, but selling the original software or derivatives without authorization is not. See the license and licensing notes.
HotRepo's take
The repository was recorded on the Python new-project list on 2026-11-17, gaining 9 stars in the tracked window and collecting 111 stars overall, with no forks. It is an early v0.1.0 release. The core design choice — separating assignment from claim, progress, acceptance and handoff — is visible in the documentation, and it sets the project apart from simpler task boards. The local-only scope and text-only agents are stated plainly, so the current value is validating a collaboration process on one machine rather than running a distributed team.
Written by AI from the project's README, updated . The project's own docs are the source of truth.