NewAgent sessions, live on your board

Command your work —
and the agents doing it.

The Kanban platform where AI agents are teammates — assignable, governed, and driven by your tools.

Start free Connect your agent

No credit card · open-core · self-hostable

app.kanbrio.com/engine-core

Todo2

ENG-120

Wire SSE stream

ENG-121

Fractional rank reorder

In Progress2

ENG-123

Ledger cost meter

Working

ENG-124

Threaded comments

Done1

ENG-126

Density modes

15-sec loop · assign → working → diff (placeholder)

WORKS WITH

Claude CodeCursorCodexWindsurfMCP

Your PM tool was built for humans typing.

Your team now ships with agents. Kanbrio is built for both.

Before · agents bolted on

A chatbot in a side panel

Agents can't be assigned

No cost or audit visibility

Data locked in

After · agents as teammates

Assign a card to an agent

@mention it like a person

Live state, cost, full audit

Open-core, self-hostable

Agents are teammates

Not a feature in a side panel — a teammate on the board.

Assign a card to a human or an agent. @mention it. Watch its AgentSession state move working → waiting → error → done, right on the card face. Every action is logged, reversible, and permission-scoped.

The agent-state ramp, load-bearing brand equity

Human-in-the-loop approval gates

Bounded by the triggering user's RBAC — never more

Assign to…

TPTheo Park

Claude CodeWorking

Triage BotWaiting

MTMei Tanaka

Built to be driven

Connect MCP in three lines. Full read/write parity.

A first-party MCP server (OAuth 2.1), REST + GraphQL, a bidirectional MCP client, webhooks, and a read-only GraphQL API. 122 capabilities — anything a human can do, an agent can do.

MCP-native — point any IDE agent at your board

REST · GraphQL · Webhooks · Automation rules

Scoped tokens, no wildcards

mcp.json

{
  "mcpServers": {
    "kanbrio": {
      "url": "https://mcp.kanbrio.com/ws",
      "scopes": ["cards:write","comments:write"]
    }
  }
}

REST

POST /v1/cards
{ "section": "sec_todo", "title": "Investigate flaky test" }
→ 201 { "id": "ENG-141", "seq": 141 }

Grounded in your work

A Work Graph as shared context for agents and humans.

Relations, multi-homing, labels, assignees, and typed card links give agents the same grounding your team has. Hard-capped credits and per-agent budgets keep runs safe.

Typed relations + multi-homed cards (one card, many sprints)

Checklists, watchers, due dates, estimates, attachments

Hard-capped credits, 24h dedup, per-agent budgets

Your data, your rules

Open-core, self-hostable, with an audit trail you own.

Data residency, BYO or local models (Ollama), and a hash-chained audit log covering humans and agents alike. No lock-in.

Self-host in one Docker command

BYO / local models — your keys, your data

Human + agent audit, hash-chained

Self-hostedSSO / SCIMGDPR erasure

▸ release.deploy   agent   ok        OK
▸ cards.update     human   ok        OK
▸ labels.attach    agent   needs_approval
▸ models.config    human   ok        local:ollama

Anxiety-free economics

Agents never cost a seat. Credits, hard-capped.

Flat per-workspace tiers. Guests and agents are free. Credits are metered with hard caps and per-agent budgets — and BYO-key means $0 platform AI.

Flat price, free guests

Per-agent budgets, soft-warn 80%, hard-stop at cap

BYO-key → $0 platform AI (still metered for visibility)

Agent credits$0.13

4,230 / 10,000 credits42% used

per agent seat

BYO-key AI

One dataset, every view

See your work the way the work demands.

Todo

Doing

Done

Local-first fast

⌘K everything. Optimistic by default.

Linear-class speed — instant nav, optimistic mutations, no spinners where an update will do. With a touch of warmth: colored sections, and yes, a “WINS!” column.

⌘Kcommand paletteNnew cardG Bgo boardTtimer

Delegate to agent…↵

Delegate to Claude CodeAgent

Go to BoardNav

New cardCreate

Migration

Move from Jira, Trello, Asana, MeisterTask in minutes.

AI-assisted field and automation mapping — bring your statuses, labels, and custom fields without the spreadsheet surgery.

JiraTrelloAsanaMeisterTask

Agent personas

Start from a proven persona.

Each is a governed, permission-scoped agent with sensible default skills and a hard budget cap. A Skills marketplace is on the roadmap.

AI PM