Fundamentals
Core Concepts
Tanjiren is a control plane for AI coding agents. Before you start using the MCP, dashboard, or desktop app, understanding these six building blocks will make everything else click.
Recommended reading order
Start with Organizations to understand access boundaries, then Workstations and Workers for the compute layer, and finally Tasks, Investigations, and Runbooks for the operational model.
Organizations
Workspace boundaries, membership, roles, and how access control flows through every object in Tanjiren.
Workstations
Registered machines running the Tanjiren desktop app. The physical compute layer where agents execute work.
Workers
The operational identity projected over a workstation: health, capacity, capabilities, and task routing.
Tasks
Units of work with intent, approval gates, execution leases, and a full lifecycle from planned to completed.
Investigations
Analytical containers that group related tasks under a shared objective, hypothesis, and findings.
Runbooks
Reusable task blueprints that encode operational patterns: intent, mode, targets, and approval requirements.
How they fit together
Organization (access boundary)
├─ Members (owner, admin, member, viewer)
├─ Security Policy (MFA, trusted controllers)
├─ Doctrine (approval rules, forbidden commands, freeze windows)
│
├─ Workstations (registered machines)
│ └─ Worker (projected identity: health, capacity, capabilities)
│
├─ Tasks (units of work)
│ ├─ Approval gate (if doctrine requires it)
│ ├─ Execution (lease-based, one worker at a time)
│ ├─ Steps (one per target workstation)
│ ├─ Logs (stdout/stderr stream)
│ └─ Artifacts (notes, links, evidence, summaries)
│
├─ Investigations (analytical containers)
│ └─ Groups related tasks + artifacts
│
└─ Runbooks (reusable blueprints)
└─ Instantiate into tasks