Jentrix documentation

Jentrix is a Trello-style task manager with five specialized board kinds (kanban, CRM, support desk, bug tracker, roadmap), built-in automations, and a first-class agent platform: every task operation in the UI is also exposed as an MCP tool, so AI agents work the same boards humans do — with their own identity, scoped tokens, and full audit attribution.

Getting started

Guide	What it covers
Getting started	Sign in, install the CLI and runner, connect Claude Code / Codex, record your first session
Set up your first project	A copy/paste coding-agent prompt that analyzes a repository, proposes a safe Jentrix topology, and creates its board, project, anchor task, and draft harness after approval
User guide	Everything in the UI: workspaces, boards & kinds, tasks, filters, saved views, bulk edit, search, My Work, inbox, dashboards, roadmap, contacts, import/export

Product features

Guide	What it covers
Projects	Project containers: boards, links, milestones, risks, and success metrics
Platform UI	Surfaces only a full deployment carries: the operate/configure/evidence rail, the Control Tower execution view, the Decision Register, Contacts, Dashboards, Releases, Harnesses, and import
Automations	The rules engine: triggers, actions, the loop guard, managing rules
Webhooks	Outbound webhooks: events, HMAC signature verification, retries, security rules
GitHub sync	Connect a GitHub App, map repos to boards, two-way Issues sync, identity mapping, loop prevention

Agent platform

Guide	What it covers
Agent platform (MCP)	The MCP server: tokens & scopes, the full tool catalog, resources, prompts, idempotency, concurrency, OAuth 2.1
MCP tool reference (product)	The product MCP catalog (the subset every deployment serves, and the MVP's whole surface): the shared tool contract and the generated per-tool table
Agent setup variants	Connecting Claude Code, claude.ai, Codex/ChatGPT, Cursor, custom SDK agents, the reusable Claude/Codex skill pack, and bundled reference agents
Agent workflows	The plan → review → execute → review loop: the agent-job substrate, the MCP tools, human gates, the `agent.handoff` webhook, and the external runner
Work orders	Bounded, versioned agent briefs: immutable Work Order versions, checksummed Context Packs, launch preview, and quick run

Governance & safety

Guide	What it covers
Policy engine	Layered, most-restrictive-wins policies that gate risky actions before they run
Credential broker	Short-lived, scoped, action-bound credential grants injected only at execution
Skill registry	Signed, versioned skills and the immutable skill pins carried by agent releases
Governed memory	Memory nomination, review, promotion, quarantine, expiry, and Context Pack projection
Communication connectors	Connector authorization, inbound normalization, policy-gated outbound delivery, and recovery
Simulation & incidents	Dry-run simulation, communication integrations, playbook mining, the incident/postmortem flow, and the MCP write-tool expansion

Software delivery

Guide	What it covers
Software Delivery OS	The delivery graph over GitHub PRs/checks + Vercel deploys: PR/CI sync, gated merge/deploy, Control Tower delivery queues, DORA + agent-native outcomes, releases, incidents, and playbook mining
Harness Automation	The governed spec → plan → stage → review → fix → validate → commit harness: the run/stage spine, reusable Runbooks, freeze/commit gates, delivery correlation, Control Tower queues, automations, the dashboard, and playbook mining
Quality & release plane	Traces, eval datasets and runs, immutable agent releases, promotion gates, and experiments

CLI

Guide	What it covers
Overview & quickstart	The `jentrix` command-line client over the MCP surface: what it is, install in one line, first commands
Install & config	Install channels (npm, npx, tarball, source), token/endpoint precedence, the config file, `jentrix login`
Command reference	The `jentrix tool` escape hatch, generated noun-verb commands, cross-cutting flags, `--json` output, exit codes
Recipes & scripting	Bash and CI patterns: idempotency keys, the CONFLICT merge loop, retries, pagination, jq pipelines
Releasing & distribution	The go-live runbook: publishing to npm, the tag-triggered release CI, verification

Operations

Guide	What it covers
Deployment	Vercel + Neon production setup: env var matrix, cron jobs, OAuth providers, optional services

Looking for product requirement documents? PRDs are planning artifacts, not user documentation — they live in prds/ in the repository.

Quick orientation

Production: https://tm.jentrix.ai
MCP endpoint: https://tm.jentrix.ai/api/mcp (Streamable HTTP, Bearer auth)
Personal access tokens: /account/tokens in the app
Repo layout: the Next.js app lives in src/; the standalone reference agents live in agents/; product requirement documents live in prds/; contributor-facing architecture notes are in CLAUDE.md / AGENTS.md

The 30-second model

Workspace  →  Board (TASKS | CRM | SUPPORT | BUGS | ROADMAP)
                 →  Column  →  Task (key like STK-123)
                                 ├─ labels, assignees, due date, priority
                                 ├─ subtasks, comments (@mentions), attachments
                                 ├─ kind sidecar: Deal / Ticket / BugReport / Initiative
                                 └─ links to other tasks (BLOCKS, DEPENDS_ON, …)

Humans mutate tasks through the UI; agents mutate them through MCP tools; automations mutate them through rules. All three paths converge on one operations core, so authorization, activity logging, notifications, webhooks, and realtime sync behave identically no matter who — or what — made the change.