sn-team-harness

SenseNova Team Harness

SenseNova Team Harness is aself-hosted team collaboration workspacethat lets people and local AI Agents collaborate around the same work: discussions, to-dos, AI execution, and final deliverables all live in one space. It connects "raising a question → assigning work → continuous progress → delivering results" into a complete, auditable collaboration chain, letting AI truly participate as a team member, rather than leaving answers in a personal chat window.

Core Concepts

Concept	Description
Workspace	The team shared space, the single authoritative source of shared collaboration facts.
Project	A collection of tasks, discussions, and deliverables under one goal.
Conversation	Team discussion; you can @Agent or create to-dos within it.
Agent	An AI participating as a team member, bound to a local Runtime to execute work.
WorkItem	A to-do item of work with an owner, status, and expected result.
Resource	Input files uploaded to a Project by people (data, materials, etc.), the raw material for Agent work, distinct from Artifacts.
Artifact	Deliverables produced by Agents (reports, code, plans, etc.), independent of chat, content-addressed, versionable, and reviewable.
Local Computer	A local client that connects services and runs Agents on members' own computers; credentials, files, and working directories all stay local.

Key Capabilities

AI enters the team's daily work: Agents participate in discussions, take on work, and reply to questions, rather than belonging only to someone's chat window.
One sentence becomes a piece of work: @Agent in a discussion or create a WorkItem, specifying the owner and expected result.
Local execution, data stays local: Agents run on members' own computers, reading Project Resources, local files, and tools; credentials and sensitive data never leave the machine.
Process and results are auditable: Task status, owners, comments, and blockers are visible; results are saved independently as Artifacts, which can be updated, downloaded, referenced, and reviewed.
Interruptions are recoverable, multiple AIs collaborate: Unpublished results are preserved during offline or concurrent conflicts; different Agents can share research, writing, programming, review, and more.

Use Cases

Research and analysis, product and operations, software development, content and design, knowledge and process management, as well as cross-role projects where multiple people and multiple AIs divide work around the same goal and people accept at key milestones.

Quick Start (Self-Hosted)

Requirements: Node.js 24 and npm. The server runs from source; no Docker image or npm package is provided.

git clone https://github.com/OpenSenseNova/SenseNova-Skills-TeamHarness.git
cd SenseNova-Skills-TeamHarness
cp .env.example .env
npm ci
npm run dev

Open in your browser http://localhost:5173. For local production mode, run npm run build then NODE_ENV=production npm start。

Basic Usage Flow

Create Workspace, invite members via a revocable Join Link.
Create Project and Agent, bind an online Local Computer Runtime。
In Conversation , @Agent, or create a WorkItem to assign an owner; upload input files as Resource to the Project.
Local Computer runs the Agent locally; the Agent reads the discussion context and Resources, replies with messages, or publishes / updates Artifact。
The team views progress, owners, and final results in the task board and Artifact version history.
When concurrent write conflicts occur, the candidate result is retained as a local Held Draft, and the Agent explicitly retries / discards / forces after reading the latest version.

Boundaries and Caveats

An early self-hosted project; it is not itself a production security boundary. Before exposing it to the public internet, add your own identity system, TLS, key rotation, backups, monitoring, rate limiting, and threat model review.
No production deployment solution, Docker image, native installer, npm package, or database migration layer is provided.
Do not commit .env, tokens, SQLite files, local working directories, or logs to the code repository.

About this skill