THE MAIN ENTRANCE
Wilco agent
Your unified coding workspace.
Read, edit, test and choose your model.
WILCO · MULTI-MODEL CODING
One coding agent.
Three model families.
One subscription.
Read project files, edit code and run checks in Wilco agent.
Pick the model yourself with /model: Claude, GPT or Grok.
Install in one line macOS · Linux
curl -fsSL https://wilco.run/install.sh | bashWindows 10 / 11: irm https://wilco.run/install.ps1 | iex · Setup guide
From $59 / 30 daysNo automatic renewal
Creating an account is free;
model usage requires a paid plan.
One shared allowance.
Official web-app memberships are not included.
Recorded sessionAvailable today
A real session, recorded 10 October 2026 with Wilco agent 20261009.2. Claude Opus 5.5 fixes a failing test; /model switches to GPT-6 Astra for a review, then to Grok 4.7 for the changelog line. Pauses were trimmed, playback is sped up, and the home directory in paths shortened to ~.
~/projects/storefrontwilco
● storefront / mainread · edit · bash
Model output is unedited. Raw recording: demo-20261010b.cast (asciinema v2).
ClaudeOpus 5.5 · Sonnet 5 · Fable 5.1 · Haiku 4.5
GPTGPT-6 Astra · GPT-5.6 Sol · Terra · Luna
GrokGrok 4.7
Catalog as of 9 October 2026. The /model picker in the agent is the live list.
BUILT FOR THE WAY DEVELOPERS WORKFiles. Tools. Terminal. Your project.
01 / Meet your agentRead. Edit. Test.
Read. Edit. Test.
In your project.
Choose a model, open your project,
and work with files and terminal tools.
YOUR PROJECT, IN CONTEXTAvailable today
Read the project. Find the context.
Work with your files, not pasted fragments from another chat.
Make the change. Run the checks.
Edit code and use terminal commands and tests in the same agent.
Choose a model for the next step.
Manually select a supported model as your work changes.
● A coding agent, with your project at the center.
02 / One subscription, your way inUse Wilco agent.
Use Wilco agent.
Or your familiar CLI.
Wilco agent is your main workspace.
Compatible CLIs use the same shared allowance.
✳
Claude Code
Compatible entry point◎
Codex
Install the CLI first↗
Grok Build
Install the CLI firstYour original logins stay separate. See setup details
These are access options, not four official memberships. Planned Wilco collaboration does not automatically control external CLIs.
03 / Where we’re heading PLANNEDDifferent perspectives.
Different perspectives.
More considered work.
A direction for Wilco—not a promise of what you can buy today.
Not available in the current release. No release date announced.
Task-aware routing
A suitable model for the task’s complexity and risk, within the choices you allow.
Independent Review
A second model checks the change. The author verifies the findings. You decide whether to enable it.
Expert Handoff
Another perspective when work gets stuck, with a deliberate handoff when it helps.
Simulated multi-model auditPLANNED
Planned workflow — not available in the current release. Play the scripted preview
~/projects/storefrontwilco
Demo complete
WILCO AGENT Claude/Verification
EXAMPLE TASK Fix cart discounts, then audit the change.
Both issues are fixed. All 12 scripted tests pass.
● Claude / verification
bash npm test -- src/cart/total.test.ts
✓ Standard, zero & excessive discounts
✓ Negative, non-finite & fractional inputs
✓ Unsafe integer boundary
Tests 12 passed · simulated result
Change ready for your review.
+ Prevented negative checkout totals.
+ Rejected invalid amounts before calculation.
↳ 2 findings resolved · rounding suggestion declined
Scripted demo complete. No code or model executed.
● storefront / mainread · edit · bash
All output is scripted. No code or model is executed.
Model roles are assigned for this example, not fixed product roles.
$ wilco · storefront / main
❯ Fix cart discounts, then audit the change.
● Claude / implementation
read src/cart/total.ts & total.test.ts
The discount is ignored. Applying it to the subtotal.
edit src/cart/total.ts
export function total(subtotal, discount) {
− return subtotal;
+ return subtotal - discount;
}
↳ Send this frozen diff for independent review.
Next: GPT, then Grok. Reviews run in sequence.
● GPT / logic review · read-only
P1 A discount above the subtotal makes the total negative.
total(2000, 2500) → −500 cents
Suggestion: clamp the result at zero.
● Grok / input review · read-only
P2 Non-finite inputs propagate into the checkout total.
total(2000, NaN) → NaN
Validate inputs and cover invalid values in tests.
Also consider rounding the result to two decimals.
↳ Return findings to the author for verification.
2 issues to investigate · 1 optional suggestion
● Claude / verify & refine
✓ P1 confirmed. Clamp excessive discounts.
✓ P2 confirmed. Reject invalid money inputs.
↳ Rounding not adopted: the API uses integer cents.
edit src/cart/total.ts
export function total(subtotal, discount) {
+ if (![subtotal, discount].every(n =>
+ Number.isSafeInteger(n) && n >= 0)) {
+ throw new RangeError("Use integer cents");
+ }
+ return Math.max(0, subtotal - discount);
}
● Claude / verification
bash npm test -- src/cart/total.test.ts
✓ Standard, zero & excessive discounts
✓ Negative, non-finite & fractional inputs
✓ Unsafe integer boundary
Tests 12 passed · simulated result
Change ready for your review.
+ Prevented negative checkout totals.
+ Rejected invalid amounts before calculation.
↳ 2 findings resolved · rounding suggestion declined
Scripted demo complete. No code or model executed.
Your models. Your permission. Your allowance. Planned controls let you pin a model or turn collaboration off. Additional routing, reviews and retries consume shared allowance. They do not guarantee savings or success. Client-side usage estimates are guidance, not enforced billing caps.
04 / Plans and allowanceChoose your plan.
Choose your plan.
Keep your choice of models.
Same agent. Same model families.
Pick the allowance that fits your work.
Try the first plan without risk. Ask within 7 days, with under 20% of the first week’s allowance used, and you get a full refund. Refund rules
Every plan includes
- Wilco agent
- Claude, GPT & Grok models
- Compatible CLI access
- One shared weekly allowance
Starter
For a project of your own.
$59/ 30 days
1,540estimated Opus prompts / month
≈ 360 / week
- Wilco agent
- Claude, GPT & Grok models
- Compatible CLI access
- One shared weekly allowance
No automatic renewal
Pro
A little more roomRoom for your daily work.
$99/ 30 days
2,570estimated Opus prompts / month
≈ 600 / week
- Wilco agent
- Claude, GPT & Grok models
- Compatible CLI access
- One shared weekly allowance
No automatic renewal
Max
For more involved projects.
$199/ 30 days
5,140estimated Opus prompts / month
≈ 1,200 / week
- Wilco agent
- Claude, GPT & Grok models
- Compatible CLI access
- One shared weekly allowance
No automatic renewal
Solo
Your largest individual pool.
$299/ 30 days
7,710estimated Opus prompts / month
≈ 1,800 / week
- Wilco agent
- Claude, GPT & Grok models
- Compatible CLI access
- One shared weekly allowance
No automatic renewal
A shared allowance
One pool. All your models.
All models and all four entry points use the same weekly allowance. It resets in seven-day windows from activation, with no rollover. Each plan lasts 30 days.
Prompt counts are estimates for Opus at a measured median task size, not guaranteed completed tasks. Monthly figures are the 30-day cap, one weekly pool for every 7 days. Long contexts and agent work use more. Future routing, reviews and retries would use extra allowance; these single-model estimates do not describe future collaborative tasks.
05 / Clear boundariesYour machine.
Your machine.
Your call.
Tools should earn your trust
by being clear about how they work.
Keep the logins you already have.
Wilco configuration is separate from your existing CLI logins. Review the install and uninstall instructions before getting started.
Local tools. Remote model requests.
The agent works on your computer. Necessary project context is sent to remote services to process model requests. Local work does not mean your code never leaves the machine.
The details are part of the product.
Read the terms, refund rules and privacy notice. Reach a person at support@wilco.run.
Official APIs, on Wilco’s own accounts.
Model requests are served through the providers’ official developer platforms, such as the Anthropic Console and the OpenAI Platform, or through OpenRouter. No shared consumer logins.
06 / A short path to your first project
Start coding with Wilco.
Install
One command on macOS, Linux or Windows.
Installing needs no account.
Create account & approve
Free. Sign in once in the browser
to approve this computer.
Choose a plan
Pick a weekly allowance when you first use a model.
30 days, no automatic renewal.
On your phone? Create your account now and continue on your computer.
Does this include official Claude, ChatGPT or Grok memberships?
No. A Wilco plan provides access to supported coding models through Wilco agent and compatible CLI entry points. Official web-app memberships and separate provider allowances are not included.
Where do the models come from?
From the providers’ official developer platforms, such as the Anthropic Console and the OpenAI Platform, or from OpenRouter, on accounts that Wilco holds. Wilco is an independent service and is not affiliated with Anthropic, OpenAI or xAI.
How is my allowance shared?
Every model and entry point draws from one weekly pool. The pool resets in seven-day windows from activation; unused allowance does not roll over. A plan lasts 30 days. Monthly prompt estimates are the 30-day cap (one weekly pool for every 7 days), not guaranteed completed tasks.
Will this replace my existing logins?
Wilco uses separate configuration and launchers, so you can keep your existing Claude Code and Codex logins. Review the setup instructions before making changes to your computer.
Which collaboration features are still planned?
Task-aware routing, Independent Review and Expert Handoff are planned and not available in the current release. The scripted terminal in the “Where we’re heading” section is an illustration; the recording at the top of the page is a real session. Current use is based on choosing a model yourself in Wilco agent. Planned collaboration is not included as a current plan benefit.
What happens when I use my weekly allowance?
Check the reset window and remaining allowance in your dashboard. You can wait for the next reset or review the available upgrade options. Usage across all models shares the same limit; switching model families does not create a separate pool. See billing documentation for the current rules.
Which platforms does Wilco support?
The current documentation lists terminal use on macOS, Linux, and Windows 10 or 11 (PowerShell). The installer sets up Wilco agent and Claude Code. Codex and Grok Build need to be installed first. Check the installation guide for current package requirements.
Where can I read the payment and refund rules?
Prices shown are in USD. Plans last 30 days and do not renew automatically. On your first plan, if you ask within 7 days and have used less than 20% of the first week’s allowance, the full amount is refunded. The existing payment and refund terms govern eligibility and requests. You can also contact support@wilco.run.
Your next project starts hereBring your project.
Bring your project.
Work in Wilco agent.
Choose Claude, GPT or Grok. Read files, edit code and run checks in one agent.
Create account Free to create an account. Plans from $59 / 30 days.