Planello Agents

Coming soon

Meet your new team.

A lead plans the card, coders build it and a reviewer checks every change, while you follow along on your board. They run on the Claude or ChatGPT plan you already have, and ask before anything you haven't allowed. Coming soon to the Planello desktop app.

Start with Planello today See what else is coming

The Checkout team's lead working on ShopFlow card #23: its chat with the planner and two coders, the team's runs in the side panel and the plan beside the chat

Preview of a feature in development. Details may change.

How it works

You give the card. The team does the rest.

Start a team on any card. The lead reads it, the planner splits it into subtasks on the card, and the work goes out to coders who build side by side. Every message they send each other shows up in Planello as it happens.

Agents

HubRunsAgentsTeams

1 thing needs youOpen

SStorefront team1 card

#28Add product search with Elasticsearch

Orchestrator o3Lead

Coder c5Coder

Reviewer r3Reviewer

CCheckout team1 card

#23Implement Stripe payment integration

Orchestrator o2Lead

Planner p2Planner

Researcher r2Researcher

Coder c3Coder

Coder c4Coder

Reviewer r2Reviewer

Orchestrator o2LeadWorking

C#23 Implement Stripe payment integration

Take #23 Implement Stripe payment integration. Card payments first; Apple Pay and Google Pay can come in a card of their own.

02:36 PM

Orchestrator o2Lead02:37 PM

On it. Planner p2 will break the card into tasks first, then I'll hand them out.

planelloget_card_details --card-id c23

Orchestrator o2Lead Planner p2Planner02:38 PM

Break #23 into subtasks on the card: the server side, the checkout form, the webhooks and the tests, in the order they block each other.

Planner p2Planner02:49 PM

Added 5 subtasks to #23: payment intent endpoint → Payment Element in checkout → webhook handler for payment_intent.succeeded → failure and retry tests → setup notes for staging.

Orchestrator o2Lead Coder c3Coder02:51 PM

Add POST /api/payments/intent: a PaymentIntent for the cart total, idempotent per cart. Tests in apps/api.

Orchestrator o2Lead Coder c4Coder02:51 PM

Add Stripe's Payment Element to the checkout page and confirm with the client secret from /api/payments/intent. Don't touch apps/api.

Waiting for Coder c3 and Coder c4 · 02:52 PM

Coder c3Coder03:30 PM

Done: POST /api/payments/intent with an idempotency key per cart, 14 tests passing. Commit a3f9c21 on planello/b1-c23-implement-stripe-payment-integration.

Orchestrator o2Lead Reviewer r2Reviewer03:31 PM

Please review Coder c3's commit a3f9c21: the amount must come from the server-side cart, never from the client.

Ask first Default effort Ultracode: Off

1. Start a team on a card

Pick the card and the team, or set a trigger so a card that moves to In Progress starts one. A whole epic works too, card by card.
2. Follow along

Each agent has its own chat. Read what they tell each other, step in with a message, or pause anyone at any time.
3. Review and merge

The work lands on the card's own branch. Try the preview, look at the changes, and merge when you're happy.

Teams and roles

Build the team that fits the work.

Start with one agent, a builder with a reviewer, or a feature team with a lead, a planner, a researcher, coders and a reviewer. Every role gets its own agent, model and instructions, and you decide who may message whom.

Mix Claude and Codex in one team: one builds, the other reviews
Run several teams at once, each on a card of its own
The team's overview shows who is doing what, with a timeline of the day

The Checkout team's overview: each agent on card #23 with what it is doing now, the one that needs you, and a timeline of the last day

You stay in charge

They ask before anything you haven't allowed.

Every agent has tool rules: what it may do on its own, what it has to ask about and what it may never do. A question waits in the agent's chat and in your inbox until you answer, and the agent can't skip it.

Coder c4CoderIdle

C#23 Implement Stripe payment integration

Orchestrator o2Lead02:51 PM

Add Stripe's Payment Element to the checkout page and confirm with the client secret from /api/payments/intent. Don't touch apps/api.

Coder c4Coder02:52 PM

Understood. I'll keep to apps/web and start a preview so you can try it.

3 steps · Read, Bash, planello

Coder c4Coder03:37 PM

The Payment Element needs Stripe's browser library. I'd like to add it to apps/web.

npm install @stripe/stripe-jsAllowed

Details

cd apps/web && npm install @stripe/stripe-js

AllowDeny

Bashnpm install @stripe/stripe-js

Coder c4Coder Orchestrator o2Lead03:39 PM

The payment form is in, on the checkout page. Try it in the preview with Stripe's test card. Ready for review.

Idle · 5 d

Waiting for the review

Review 1

Checkout page, with the payment form

http://localhost:5173/checkout

Use the card 4242 4242 4242 4242 with any future date.

Open in PlanelloOpen in browser

Running 1

npm run dev (apps/web)

Command

Plan 4

Read the checkout page

Add the Payment Element

Confirm the payment and show errors

Screenshot in light and dark

Ask first
Accept edits
Auto
Full auto
Tool rules per agent
One inbox for every question

Everything a team needs

Built for real work on a real project.

Agents work in your project folder with your git, and on your board like you do: they move cards, tick off subtasks and write documents. Everything they learn stays in your workspace.

Teams

Ready-made teams to start from, with plain names you can change.

One agent
Builder and reviewer
Feature team
Who talks to whom
Team chat

Git

Each card gets a branch and a folder of its own, so your checkout is never touched.

A worktree per card
Commits named by agent
You approve each merge
Or pull requests

On the board

Agents use your board the way Planello AI does, and claim the card they're on by moving it.

Move cards
Add and tick subtasks
Write documents
Work through epics

Memory and instructions

What every agent should know, and what each one learns, in notes you can read and edit.

Workspace instructions
Agent memory
Skills
Changes they propose

Automation

Triggers act when cards change, so work starts without anyone asking.

Start a team when a card moves
Ask the reviewer at review
Call a webhook
Notify me

Following along

See every agent at once, or one in detail, without leaving Planello.

Agents Hub
Org chart
Timeline
Previews in Planello
Commits by card

Agents Hub

Everything your agents do, in one place.

The Hub opens on what needs you, what is running and what was done lately, next to your plans' limits for the week. Start a team, an epic or a chat from there.

Agents Hub: one tool waiting to be allowed, this week's cost and plan limits, two teams running and recently finished work

Coder c4's chat asking to install a package, with the checkout preview, the running dev server and its plan beside it

Usage

Know what every card cost.

Planello counts the tokens of every agent, team and run, as the AI reports them, and shows what the same use would cost at API prices. Here is the Checkout team's last two weeks: the shopping cart, then Stripe.

Checkout team

6 seats

OverviewUsageRoles 5InstructionsMessagingCard work

7 days30 days90 daysAll time New tokensAll tokensCostTurns

Cost at API prices: $45.56; Estimated from current API prices. Your subscription covers it, within its limits.
New tokens: 5.3M; 37.8M with cached input
Turns: 413
Runs and chats: 2
Agents at work: 12

Tokens by kind

New input1.5M4%
Cached input32.5M86%
Cache writes2.3M6%
Output1.5M4%

Cost per day

$0.00$2.00$4.00

Oct 10Oct 12Oct 14Oct 16Oct 18Oct 20Oct 22

By agent

C CoderCoder 68% 3.6M new
25.6M in all $30.90
R ReviewerReviewer 13% 662.3K new
4.7M in all $5.70
O OrchestratorLead 10% 541.5K new
3.9M in all $4.66
P PlannerPlanner 7% 395.1K new
2.8M in all $3.40
R ResearcherResearcher 2% 104.6K new
747K in all $0.90

Per agent
Per team
Per card and run
Cost or tokens
Cached input apart

Your plans

Runs on the plan you already pay for.

Agents run on your own Claude or ChatGPT subscription, or on an API key when a plan's limits aren't enough. Planello shows each plan's limits while the team works, so a busy afternoon doesn't use up your week.

The 5-hour and weekly limits, and how many agents use each plan right now
Each role gets a fitting model: the strongest to plan, a quick one to look things up
Local models in LM Studio or Ollama run on your own computer

This week

Cost at API prices: $32.11
New tokens: 3.7M
Turns: 290

Claude Agent · Max6 running

5-hour window38%

This week61%

Codex · Pro2 running

This week22%

Safe by design

Your project stays yours.

Agents work on your computer, in your project, under rules you set. Nothing reaches your main branch until you say so.

Ask first

Commands and file changes you haven't allowed wait for your answer. In the ready-made teams, reviewers and leads don't edit files.

A folder per card

Each card's work happens on its own branch, in its own folder, so your checkout is never touched.

You merge

Agents ask when the work is ready. You look at the changes and merge from Planello, or they open a pull request.

On your computer

Agents run in the desktop app. Your code goes only to the AI you chose, and with a local model it never leaves your computer.

Works with your AI

Claude, Codex or a model of your own.

Each agent runs on the AI you pick for it. Use your Claude or ChatGPT plan, an API key, or any model from AI Configuration, including local ones, which Planello gives files, search, a shell and the board.

A model for each agent, and how hard it thinks, from low effort to max
Switch an agent to another AI at any time; its team stays as it is
Agents get the same board tools as Planello AI, so they know their way around

ClaudeYour Claude plan or an Anthropic API key
CodexYour ChatGPT plan or an OpenAI API key
Local modelsLM Studio or Ollama, on your computer
Any providerAny model from AI Configuration

Questions

Good to know

What do I need to run agents?

The Planello desktop app and a project folder, plus one way for the agents to think: a Claude or ChatGPT subscription, an API key, or a local model in LM Studio or Ollama. The setup walks you through it.

Does it work in the browser?

No. Agents need your project folder, git and a terminal, so they run in the desktop app. Your boards are the same in both, so cards you plan in the browser are ready for them.

Can agents break my project?

They work on the card's own branch in a folder of their own, they ask before anything you haven't allowed, and their work reaches your main branch only when you merge it. You can pause or stop any of them at any time.

What does it cost?

Planello itself is free. Agents use your own AI: on a subscription they count toward your plan's limits, on an API key you pay the provider. Planello shows what the same use would cost at API prices, so you can compare.

Do agents only write code?

Code is what they do best, but a researcher can look into a question and leave a document on the card, and every agent can plan cards, add subtasks and keep the board up to date.

When can I use it?

Planello Agents is in development and coming soon to the desktop app. Planello is free, so you can start planning today: when agents arrive, your boards and cards are ready for them.