The landing-page system for agents

Your agent can build the page. LP1 runs the marketing loop.

Create and iterate Page Sets, measure every view and conversion, spot the pages that break away, and automate the next test—without rebuilding hosting, forms, webhooks, or analytics.

Start building See the operating loop

Bring your OpenRouter key or bring the agent you already use.

Summer launch pages

12 Pages · live Page Set

Measuring

Views

18,492

Conversions

1,126

Rate

6.1%

Direct comparison page

Breaking away from the set baseline with enough traffic to investigate the pattern.

+34%

Agent recommendation ready

Review evidence →

Build

Give the job to an agent

Use LP1 Agent with your own OpenRouter key, or connect Codex, Claude Code, Cursor, or a supported ChatGPT workspace app through LP1 MCP.

Learn

See the whole Page Set

LP1 keeps views, form impressions, conversions, experiment arms, and outlier evidence together so the next move is obvious.

Operate

Automate inside your rules

Let the agent observe, prepare drafts, start tests, or publish qualified winners inside budgets, cooldowns, stop-loss rules, and explicit owner approval.

See the whole set

Stop opening twelve analytics tabs to understand twelve pages.

Page Sets put the Pages for one audience, offer, or campaign in the same operating view. Views, form impressions, conversions, and rates stay attached to the exact Page Versions that earned them.

Page performance

Comparison page

5,208 views

8.4%

Pain-first page

4,916 views

6.2%

Founder story

4,601 views

4.9%

Find the pages that matter

LP1 points to the outlier—and shows its work.

A page is not called a winner because it got three lucky clicks. LP1 waits for the owner’s minimum sample, compares performance to the Page Set baseline, and explains the evidence an agent is allowed to act on.

Outlier evidence

Outperforming

Comparison page

1,492 qualified views. Conversion rate is 34% above the Page Set baseline with a score above the investigation threshold.

Sample: 1,492
Baseline: 6.1%
Page: 8.2%

Run controlled experiments

Versions become arms. Visitors stay in the arm they were given.

Create weighted experiments from immutable Page Versions. LP1 keeps assignment stable, records every event against the right arm, watches sample balance, and pauses tests when a stop-loss is crossed.

Hero direction

Running

Control · Version 18

50% · 5.8%

Agent variant · Version 21

50% · 7.4%

Stable assignment · balanced sample · no stop-loss triggered

Let the agent operate

Automation is a ladder, not an on switch.

Start in observe mode. Move to draft or test when the evidence is trustworthy. Enable publish only when the owner is ready. Daily budgets, cooldowns, sample gates, experiment limits, and stop-losses remain in force at every step.

Automation policy

Observe Draft Test Publish

Daily changes: 2
Minimum sample: 500
Cooldown: 24 hours
Stop loss: 20%

Latest run started one prepared experiment. Every other Page stayed unchanged.

Keep the marketing plumbing

The form, conversion, webhook, and delivery receipt are already here.

LP1 detects managed forms, stores accepted submissions, records the conversion atomically, signs outbound events, retries transient failures, and keeps delivery history and replay close to the Page Set.

Delivery history

CRM intake

qualified_lead

Delivered

Lead alert

managed_form

Delivered

Warehouse

page_view

Retrying

Form capture succeeded. One external endpoint is retrying in the background.

Two ways to work

Use LP1 Agent. Or bring yours.

LP1 Agent

Your OpenRouter key, saved for your account.

Choose LP1’s fast, balanced, or best preset. The credential is encrypted, the full key is never shown again, and Page Chat stays unavailable until the owner configures it.

Your existing agent

Codex, Claude Code, Cursor, or a supported ChatGPT workspace app.

Connect through LP1’s hosted MCP server. The owner approves separate capabilities for Page access, publishing, marketing data, experiments, automation, and automated publishing.

Owner control

The agent gets a job. The owner keeps the boundary.

Every external connection can be reviewed and revoked. Consequential work leaves an activity record. Failed delivery does not erase a captured lead. Publish automation cannot turn itself on.

Credentials: Encrypted at rest; previews only
Agent access: Capability-scoped Owner API Tokens
Automation: Budgets, cooldowns, stop-loss, audit trail
Webhooks: Signed, retried, visible, replayable

Build the next Page Set

Let the agent make pages.
Let LP1 run the loop.

Start building