A terminal coding agent and Mac app, like Claude Code and Codex — not an IDE. It works in your repo, alongside your editor.

Half price, similar or more work done.

Same frontier models (Anthropic, OpenAI, xAI) as the $20–$200 plans you already know — at half the price. Light is $10, Pro is $30, Heavy is $100.

Half the price
$10, $30 and $100 against the $20, $60 and $200 tiers you know. Same models.
Frontier models included
Anthropic, OpenAI and xAI through our gateway, on our keys. You never paste a provider key or see a token bill.
No restarts
Work all day in one session — no "start a new chat", no re-explaining what you already said.
Install · Terminal
$ curl -sSL https://storage.googleapis.com/assertion-downloads/install.sh | sh
$ cd your-repo && assertion
Download the Mac app Apple silicon · .dmg

One checksum-verified assertion binary in /usr/local/bin; assertion upgrade keeps it current.

Why half price

How we get to half price with the same or more work done

Normal agents resend the whole conversation on every step, so the cost per step keeps climbing all day. Ours keeps the conversation in a fixed-size window and stores the rest outside the prompt, so the cost per step stays flat. We pass the difference on to you.

Input tokens per step, as the session goes on
Assertion AgentA normal agent
Input tokens per step across a 24-step task: a normal agent climbs from 11K to 158K; Assertion stays between 11K and 27K.

Measured on a real 24-step build task run twice with the same model: once on a normal agent, once on Assertion. Same prompts, same tests passing at the end.

Pricing

Same frontier models. Half the price.

You never see a token bill. Each plan includes a week of usage, shown as a percentage that resets every week.

PlanPriceIncluded
Free$0About 15–20 prompts a week. Enough to see it work, not enough to live on.
Light$10/moA working week for one person.
Pro$30/mo3× Light. All-day use, the larger models when you want them.
Heavy$100/moFor people who run the agent all day, every day.

For reference, the $20 tier elsewhere: Cursor Pro $20 · Claude Code Pro $20 · ChatGPT Plus with Codex $20. Light is $10. List prices, September 2026.

Teams

One plan per team. Buy seats — $20 a standard seat, $60 a premium seat (5× the allowance) — and hand them to people; only seat-holders draw on the plan. Usage is pooled across the team, so a quiet week for one person is slack for another.

What "included usage" means

Real compute, metered honestly — a larger model draws the week down faster, and the picker says so where you choose.

Security

Your code goes to the model, as with any agent, and nowhere else.

Model calls

Go through Assertion's gateway to Anthropic, OpenAI or xAI on our keys — the same providers behind the plans you use today. You never hold or paste a provider key.

What we store

Your account, your plan and your usage, plus the agent's own working notes for your sessions, private to you. Not your repository.

Assurance

SOC 2 audit in progress. Independent penetration test completed, with no critical, high or medium findings.

Teams

Owners buy seats and hand them to people. Only seat-holders draw on the plan, and owners manage members and seats.

Where this is

It's an alpha, and we'd rather say so.

macOS, Apple silicon only

No Intel build, no Windows or Linux app yet. The CLI installer is Apple silicon too.

What's rough

It's new. Expect sharp edges, and tell us at support@assertion-ai.com when you hit one.

Also a plugin

Already on Claude Code, Codex or Cursor? We ship a plugin for those too — install it here. The half price is the agent's, not the plugin's.

Questions
Can I use my own model keys?

Not in the alpha. Models are included in the plan and go through our gateway; that's how usage stays a percentage instead of a bill.

What if I hit the week's limit?

The agent tells you plainly, with the reset day and an upgrade link. Nothing is lost; the week resets and you carry on.

How does this compare to Cursor Pro?

On price: Light is $10 to Cursor Pro's $20, Pro is $30 to Pro+'s $60, Heavy is $100 to Ultra's $200, on the same frontier models (list prices, September 2026). On form: this is a terminal agent and a Mac app, not an IDE — so the closer comparison is Claude Code or Codex, and against those the difference is the price.

Does it work with a team?

Yes. On a team plan the owner buys seats, standard or premium, and hands them to people. Everyone's agent draws on one pooled allowance, billed per seat.

Windows or Linux?

Not yet. The installer and the Mac app are macOS on Apple silicon for now.