Features — Zooid, the coding agent you would build yourself

01 · WORKSPACE

Every project. One calm window.

Run agents on several repos at once without switching apps or losing your focus. Each agents works in isolation and pings you only when it needs a decision.

billing-apichange/add-ratelimitAdd rate limiting to /invoices#241

Agent

Add rate limiting to /invoices: 100 requests per minute per API key. Return a 429 with Retry-After, and add tests.

Done. Requests to /invoices are now limited per API key with a sliding window, and over-limit calls get a 429 with a Retry-After header.

Createdsrc/middleware/rateLimit.ts+34Editedsrc/api/invoices.ts+3 −1Rannpm test -- invoicesexit 0 · 2.4s

All 18 tests pass, including two new ones for the 429 path. I opened a draft PR so CI can run; tell me if you want it marked ready for review.

Worked for 3m 12s

Ask the agent to change something, @mention files, run /commands

Sonnet 4.5High

Draftnot asking for review yetReady for review

OverviewReviewChanges4FilesTerminalPreview

Plan3 of 4

Add sliding-window limiter middleware

Wire it into the /invoices routes

Tests for limit, reset and 429 response

Wait for CI on the draft PR

Details

ModelSonnet 4.5 · High

Branchchange/add-ratelimit

Context42k of 200k

Spent$0.14

Every project, one sidebar
Every repo you work on lives in one sidebar, running agents included. Jump between them in a click, without juggling windows.
A worktree per agent
Every session gets its own git worktree and branch, so parallel agents never collide with each other or with your checkout.
Pick up where you left off
Sessions keep their chat, plan, diff and context. Close the app, come back next week, and type “continue”.
Desktop or terminal
Live in the shell? The Zooid CLI puts the same agent, with every capability, in a fast terminal UI.

Start from main, a PR, a branch or an issue

Branch off main, or pick up an open pull request, an existing branch or a GitHub issue. The agent starts with that context already loaded.

One inbox for everything that needs you

Reviews, command approvals, failing checks and stalled model calls from every project land in Needs you, with one-click actions.

02 · THE AGENT

It plans, delegates, and checks its work.

Under both the desktop app and the CLI runs the same agent harness. It splits big jobs across subagents, runs on its own when you let it, and verifies changes before it says it's done.

❯/new

── new session 7KQ2M9FXa1 ──

❯spawn 3 subagents to audit the billing api

OVERVIEWctrl+t

cwd~/code/acme/billing-api

context▱▱▱▱▱ 18.7k / 1.0M · 1%

session7KQ2M9FXa1

AGENTS●3 running · /kill <id>

●mainrunning job…30s

├●a1exploreMap every route that touches invoices16s

│round 6/8 · grep router\.(get|post)

├●a2exploreAudit auth on the billing endpoints16s

│round 7/8 · read src/auth/apiKey.ts

└●a3exploreFind missing tests for rateLimit12s

round 4/8 · read tests/invoices.test.ts

01
Built-in harness + subagents
The main agent plans the job and hands parts of it to subagents that explore, audit and test in parallel. You see every one of them, and can stop any of them.
02
A plan you can follow
Every task turns into a live checklist, with the model, context and spend right beside it. You always know which step it's on.
03
Auto mode, with a Safe mode
Let it run without approving every command. Safe mode checks each action against your request and stops for anything risky, so it never goes off the rails.
04
Tests your app like a user
With built-in browser tools it opens your web app, clicks through it, takes screenshots and checks the result before it calls a change done.

Skills
Teach it how your team works once. Skills sync into every session.
MCP connections
Plug in any MCP server: your tracker, error monitoring, docs or database.
Sees what you see
Drop in a screenshot or a screen recording and it becomes context.
Your model, your key
Pick the model and effort per session, or bring your own API key.
Context & cost
Live context use, automatic compaction, and the exact spend per session.

03 · BUILT-IN TOOLS

No more alt-tab. It's all in the app.

Terminal, browser preview, file explorer and diff live in the same window as the agent, all pointed at the session's own worktree.

A browser preview, next to the code

Your dev server runs in a Preview tab beside the agent. Every change shows up live, without leaving the app.

A real terminal, in the worktree

Open a shell right inside the session's worktree to poke around, run a script or double-check the agent.

Point at the exact lines

Select code in any file, press ⌘L, and those lines drop into the chat as context. No copy-pasting paths.

billing-apiAdd rate limiting to /invoices

AgentAll changesREADME.md

README.mdModified+12RenderedSource

billing-api

Invoicing and payments API for the Acme platform. Node 20, Express and Postgres.

Run locally

npm install && npm run dev

Rate limits

RouteLimit

/invoices100 / min per key

/customers300 / min per key

/webhooksNo limit

Go to file⌘P

src

api

invoices.tsM

customers.ts

middleware

rateLimit.tsA

server.ts

package.json

README.mdM

Files, with rich previews

Browse the tree, jump to any file with ⌘P, and read READMEs and docs rendered, or flip to the source.

CHANGES VIEWER

Every change in one diff

All changes against main in one scrolling view, with the file list beside it. Switch between unified and split, revert a hunk, and tick files off as you go.

Unified, split and file view
Mark files as viewed
Open any file in its own tab

billing-apiAdd rate limiting to /invoices#241

AgentAll changes

src/middleware/rateLimit.ts+34Viewed

1import { Request, Response, NextFunction } from 'express';

2

3const WINDOW_MS = 60_000;

4const LIMIT = 100;

5const hits = new Map<string, number[]>();

6

7export function rateLimit(req, res, next) {

8const key = req.header('x-api-key');

src/api/invoices.ts+3−1Viewed

12router.get('/invoices', listInvoices);

12router.get('/invoices', rateLimit, listInvoices);

13router.post('/invoices', rateLimit, createInvoice);

OverviewChanges 4Files

4 files changedagainst main · +52 −1

ArateLimit.ts+34

Minvoices.ts+3 −1

ArateLimit.test.ts+11

MREADME.md+4

GITHUB REVIEW

Comment, and it lands on GitHub

Leave a review comment on any line. It posts to the pull request as you, so your team reads the same thread you do.

Line comments sync to the PR
Draft → Ready for review in one click

PRS, CHECKS & AUTOFIX

Red check? Hit Autofix.

Open the pull request from the app and watch CI run right beside the diff. When a check fails, Autofix hands the log to the agent and it pushes the fix.

Issue → session → pull request
Live CI status per check
Autofix from the failing log

billing-apichange/add-ratelimitAdd rate limiting to /invoices#241

AgentAll changes

TSsrc/middleware/rateLimit.ts+38Viewed

1import type { Request, Response } from 'express';

3const LIMIT = 60; // per API key, per minute

4const buckets = new Map<string, Bucket>();

6export function rateLimit(req, res, next) {

7const key = req.header('x-api-key');

TSsrc/api/invoices.ts+3−1Viewed

12import { listInvoices } from './handlers';

13router.get('/invoices', listInvoices);

13import { rateLimit } from '../middleware/rateLimit';

14router.get('/invoices', rateLimit, listInvoices);

DraftReady for review

OverviewReviewChanges 4

Add rate limiting to /invoiceschange/add-ratelimit → main · 2m ago

Checks1 failed· 1 passed

Install & typecheck41s

Test · invoices1m 2s

✕ invoices › returns 429 after 60 requests

AutofixView log

SAFE MODE · ON BY DEFAULT

Almost autopilot.
Never reckless.

Safe Mode lets the agent work through a task on its own, using only commands that can't harm your machine, your repo or your data. A riskier step is only proposed when it's really needed, and then the agent stops and asks you first.

Reads, builds and tests run without approval prompts
Every action is checked against what you asked for
Risky commands only when necessary, never without your OK

agent · feat/billing-exportSAFE MODE

git statusran safely

rg "exportCsv" src/ran safely

pnpm installran safely

rm -rf ~AUTO-DENIED

Deletes your home directory, outside the workspace. Blocked by policy, no prompt needed. The agent carried on with the task.

pnpm test --filter billingran safely

pnpm lint --fixran safely

git push --force-with-leaseNEEDS YOUR OK

Needed to update the PR after the rebase you asked for. This rewrites history on the remote branch.

ApproveReject

PRIVACY, BUILT IN

Stored on your machine
Sessions, history and code live on your own machine, not on our servers.
Zero data retention
Inference runs in memory. Prompts and code are never stored by the model provider.
Never used for training
Your prompts and code never train a model. Not ours, not the model provider's.
EU, end to end
Models run only on EU providers, and Zooid is a Belgian company with no US parent.

06 · COMING NEXT

Want to see what's coming next?

The features our team is building right now. Pilot users help decide the order, and get each one first.

IN DEVELOPMENT

Cloud sandboxes

Run agents in secure, isolated sandboxes in the EU instead of on your own machine.

IN DEVELOPMENT

Pipeline builder

Chain agents into workflows that take a ticket all the way to a reviewed pull request.

DESIGNING

Multiplayer coding

Work alongside your teammates and their agents in the same session, in real time.

DESIGNING

Mobile app

Check in on your agents from your phone: approve plans and review pull requests on the go.

+ more on the roadmap, all the way to a full SDLC platform.

See the full roadmap

Try all of it.
Free in the beta.

Every shipped feature on this page is in the beta today. Request an invite and put it to work on real code.

Request an invite

EVERYTHING ZOOID DOES,

Every project. One calm window.

Every project, one sidebar

A worktree per agent

Pick up where you left off

Desktop or terminal

Start from main, a PR, a branch or an issue

One inbox for everything that needs you

It plans, delegates, and checks its work.

Built-in harness + subagents

A plan you can follow

Auto mode, with a Safe mode

Tests your app like a user

Skills

MCP connections

Sees what you see

Your model, your key

Context & cost

No more alt-tab. It's all in the app.

A browser preview, next to the code

A real terminal, in the worktree

Point at the exact lines

Files, with rich previews

From issue to merged, without leaving Zooid.

Every change in one diff

Comment, and it lands on GitHub

Red check? Hit Autofix.

Let it run.
With peace in mind.

Almost autopilot.
Never reckless.

Stored on your machine

Zero data retention

Never used for training

EU, end to end

Want to see what's coming next?

Cloud sandboxes

Pipeline builder

Multiplayer coding

Mobile app

Try all of it.
Free in the beta.

EVERYTHING ZOOID DOES,

Every project. One calm window.

Every project, one sidebar

A worktree per agent

Pick up where you left off

Desktop or terminal

Start from main, a PR, a branch or an issue

One inbox for everything that needs you

It plans, delegates, and checks its work.

Built-in harness + subagents

A plan you can follow

Auto mode, with a Safe mode

Tests your app like a user

Skills

MCP connections

Sees what you see

Your model, your key

Context & cost

No more alt-tab. It's all in the app.

A browser preview, next to the code

A real terminal, in the worktree

Point at the exact lines

Files, with rich previews

From issue to merged, without leaving Zooid.

Every change in one diff

Comment, and it lands on GitHub

Red check? Hit Autofix.

Let it run. With peace in mind.

Almost autopilot. Never reckless.

Stored on your machine

Zero data retention

Never used for training

EU, end to end

Want to see what's coming next?

Cloud sandboxes

Pipeline builder

Multiplayer coding

Mobile app

Try all of it. Free in the beta.

Let it run.
With peace in mind.

Almost autopilot.
Never reckless.

Try all of it.
Free in the beta.