MultiOn pricing

About

MultiOn was one of the first autonomous web-agent companies: AI agents that operated a real browser to complete multi-step tasks — search, click, fill forms, check out — from a natural-language command. Founded in 2022 by Stanford PhD dropout Div Garg and MBA classmate Omar Shaya, it shipped a consumer Chrome extension and mobile app (“Your Personal AI Agent”), then launched a developer-facing Agent API in public beta in April 2024, billed per request. Backers included General Catalyst, Amazon’s Alexa Fund, Samsung Next and Maven Ventures; a $20M Series A led by General Catalyst closed in June 2024 at a reported ~$100M valuation, and the much-discussed Agent Q research (self-healing agents with planning) landed in August 2024.

It didn’t last. In its December 20, 2024 end-of-year update, MultiOn announced a shift to a “purely consumer-focused product & technology company” with a rebrand to follow — it became Please (please.ai) — and co-founder Div Garg departed to found AGI Inc (theagi.company), an applied agents research lab. By mid-February 2025, multion.ai had stopped resolving and began 301-redirecting to theagi.company, whose /pricing URL returns a 404. The Agent API and its public price card are gone; this page documents the last-known pricing as a post-mortem.

For the historical record, see the archived MultiOn API pricing page.

Pricing summary : How MultiOn’s pricing model worked

MultiOn’s Agent API was pure usage-based pricing on a single meter: the request, defined as one action the agent takes on a webpage. The pricing page spelled the unit out with an example — buying a book on Amazon takes about 5 requests (search, click the link, add to cart, check out, purchase) — making it one of the most legible agent-pricing explanations of the 2024 cohort.

The last-known structure (Wayback, 2024-11-23) had four rows: Basic at $0.04/request (0–999,999 requests, no minimum, self-serve), Premium at $0.025/request (2M–9,999,999 requests, $50,000/year minimum spend), Retrieve API at $0.01/request (extraction-only calls, any volume), and Custom (“get in touch”) above 10 million requests. There was no free tier and no subscription — the consumer app monetized separately, and the API monetized purely on metered actions plus an annual-commit discount tier.

What makes this different: MultiOn priced the action, not the task, the token, or the minute. That made unit economics transparent (every click had a price) but meant task costs scaled with how many steps the agent needed — and the agent, not the buyer, decided the step count. The model also moved fast: prices were halved within about five weeks of public beta, an early signal of how steep the cost-deflation curve under agent APIs would be.

Pricing by product

Sales motions across products: self-serve signup for Basic and the Retrieve API; sales-assisted for the $50K-minimum Premium tier; sales-led for Custom. All rates above are historical (last archived 2024-11-23) — the product no longer exists.

Hidden costs : What MultiOn users actually paid

The per-request rate looked tiny, but tasks consumed multiple requests and the agent decided how many. MultiOn’s own example — a 5-request Amazon purchase — meant a completed task cost $0.20 at the $0.04 rate ($0.40 at launch pricing). Agents that retried, navigated bot walls, or wandered through extra pages burned extra billable actions with no published cap per task.

Two structural catches: the Premium minimum ($50K/year) made the discounted rate a commitment product, not a volume discount you grew into automatically; and after the May 2024 re-banding, the published tiers skipped the 1M–2M request range entirely — a buyer there paid Basic unit rates that could exceed Premium’s full annual minimum. “Prices subject to change” was printed on the card, and MultiOn used it: rates halved roughly five weeks after launch.

Want to estimate your own MultiOn-style bill? Use the MultiOn pricing calculator to model per-request agent costs based on usage patterns.

Pricing evolution : MultiOn pricing history and changes

Cadence

Tracked range: 2023–2025, via Wayback Machine snapshots of multion.ai/pricing (2023-12-03) and multion.ai/api/pricing (2024-04-14 through 2024-11-23). No live page exists to verify today.

Notable changes

• 2024-04 — Agent API enters public beta with Basic $0.08/request (no minimum), Premium $0.05/request (1M+ requests, $50K/yr minimum), and Custom above 10M requests.
• 2024-05 — Prices halved: Basic $0.04, Premium $0.025. Premium’s entry point moves from 1M to 2M requests, stranding the 1M–2M band on Basic rates.
• 2024-08 — Retrieve API appears at $0.01/request (any volume) for extraction-only calls; Agent Q research is published the same month.
• 2024-12-20 — EOY update announces the consumer pivot and rebrand (to Please); Div Garg departs. API pricing last archived unchanged on 2024-11-23.
• 2025-02 — multion.ai goes dark, 301-redirecting to theagi.company (AGI Inc). No successor public price card exists; please.ai carries no per-request API pricing.

What’s unique : MultiOn’s distinctive pricing mechanics

1. Priced the action, and said so in plain English. Most 2024 agent APIs priced tokens, minutes, or vague “credits”. MultiOn priced the webpage action and explained the unit with a concrete 5-request shopping example right on the price card — unusually honest mechanics for the category.

2. A 50% cut five weeks into public beta. Launching at $0.08 and halving to $0.04 by late May 2024 is one of the fastest documented public price cuts in the agent cohort — part land-grab, part tracking the collapsing inference costs underneath.

3. Commit-gated discount with a dead zone. The cheaper $0.025 rate wasn’t a volume discount you aged into — it required a $50,000/year minimum and (after May 2024) at least 2M requests, leaving a published gap where 1M–2M-request customers paid more at Basic rates than Premium’s entire annual minimum.

Strengths & weaknesses

Billing UX : MultiOn billing controls and transparency

• Billing controls — Basic and the Retrieve API were self-serve: sign up at app.multion.ai, get keys, pay per request with no minimum. Premium and Custom went through sales (“get in touch”) with annual minimums. Historical — the signup flow no longer exists.
• Usage visibility — The public price card was unusually clear about the unit (action-level metering with a worked example), but there was no published per-task estimator, so forecasting a bill required guessing the agent’s average step count per task.
• Payment options — Standard self-serve API billing for Basic/Retrieve; invoiced annual-commit contracts for Premium ($50K/yr minimum) and Custom. Details beyond the price card are not recoverable from the archive.

Strategic wins : Why MultiOn’s pricing decisions worked

1. The action meter made an opaque product legible

Agent workloads are hard to price because the work varies per task. Metering the atomic action — and printing a worked example — gave developers a unit they could instrument and audit. It remains one of the clearest statements of agent unit economics from the 2024 cohort. See choosing the right usage metric.

2. Cutting price into the land-grab

Halving rates five weeks into public beta tracked collapsing model costs and undercut per-task competitors while the category was still forming. For a market where the buyer fear was runaway metered bills, leading the price down was the right direction of travel. Related: token cost deflation and AI pricing.

3. Splitting read from act

The $0.01 Retrieve API recognized that extraction-only calls cost less to serve and compete with scrapers, while full action-taking justified 4x the rate. Tiering by capability consumed — rather than one blended rate — is a pattern later agent platforms adopted. See usage-based pricing strategy.

Areas to improve : Gaps in MultiOn’s pricing approach

1. The buyer didn’t control the meter

Charging per action when the agent chooses the actions means cost variance the customer can’t manage: retries, dead ends, and bot walls all billed the same as progress. Without a per-task cap or outcome-based pricing, heavy automation carried unbounded downside. See bill shock and cost unpredictability.

2. A published tier gap

After May 2024, Basic ended at 999,999 requests and Premium began at 2,000,000 with a $50K minimum. The 1M–2M band — arguably the natural growth path for a successful Basic customer — matched no listed tier and cost more at Basic rates than Premium’s whole annual minimum. That’s a packaging error, not a strategy.

3. Pricing outlived the strategy, not the company

The price card never changed after August 2024 even as the company pivoted underneath it; the API was wound down within nine months of public beta. Developers who built on the $0.04 meter had no published migration or deprecation pricing — the page simply vanished into a redirect. Post-mortem lesson: agent-API buyers should weigh vendor durability as part of the price.

Key takeaways

1. Price the unit your buyer can audit. “1 request = 1 webpage action,” with a worked example, is still a model explanation of agent metering — clearer than credits or blended task rates.
2. Expect violent early repricing in new categories. A 50% cut five weeks after launch shows how unsettled agent economics were in 2024 — early list prices are opening bids, not anchors.
3. Don’t publish tier gaps. MultiOn’s 1M–2M request dead zone made its own discount tier irrational for a real customer segment.
4. Usage meters need cost-control surfaces. When the agent decides the step count, buyers need caps, estimators, or per-task pricing to trust the meter.
5. A transparent price card is not a durable business. MultiOn had the most legible pricing in its cohort and still wound the API down within nine months — pricing quality can’t outrun strategy churn.

UBP implications

1. Action-level metering is the most auditable agent unit, but it transfers variance to the buyer. Pure per-action pricing works only with cost-control tooling (caps, previews, success filters) the 2024 cohort — MultiOn included — didn’t ship.
2. Capability-tiered rates (act vs. retrieve) beat blended rates. Charging 4x for action-taking over extraction matched price to cost-to-serve and let one API serve two budgets.
3. Commit-gated discounts need continuous tier coverage. If the discounted tier starts above where the standard tier ends, the gap customers are mathematically better off churning — check the bands before publishing. See usage-based pricing fundamentals.

Sources

• Archived MultiOn Agent API pricing — Wayback 2024-04-14 ($0.08/$0.05) (accessed 2026-06-10)
• Archived MultiOn Agent API pricing — Wayback 2024-05-21 ($0.04/$0.025) (accessed 2026-06-10)
• Archived MultiOn Agent API pricing — Wayback 2024-11-23 (last snapshot, +Retrieve $0.01) (accessed 2026-06-10)
• MultiOn EOY Team and Business Updates — Wayback 2024-12-20 (consumer pivot, rebrand, Div Garg departure) (accessed 2026-06-10)
• MarkTechPost: MultiOn releases Agent API in public beta (April 2024) (accessed 2026-06-10)
• Synthedia: MultiOn’s nine-figure valuation ($20M Series A, General Catalyst) (accessed 2026-06-10)
• Amazon Alexa Fund on MultiOn at CES 2024 (accessed 2026-06-10)
• Dealroom: Please, formerly MultiOn (accessed 2026-06-10)
• Cerebral Valley: AGI, Inc. — from MultiOn’s Div Garg (accessed 2026-06-10)
• Liveness checks (2026-06-10): multion.ai and multion.ai/pricing 301-redirect to theagi.company; theagi.company/pricing returns 404; please.ai resolves as the consumer successor.

Bottom line

MultiOn was a pioneering autonomous web-agent startup whose Agent API priced the atomic unit of agent work — $0.08 per webpage action at its April 2024 public beta, halved to $0.04 within five weeks, with a $0.025 commit-gated Premium tier ($50K/yr minimum) and a $0.01 extraction-only Retrieve API. Despite a $20M General Catalyst-led Series A and the Agent Q research moment, MultiOn announced a pivot to a purely consumer company in December 2024, rebranded to Please, and its founder left to start AGI Inc — multion.ai now redirects to theagi.company and the price card survives only in the Wayback Machine. This page is a post-mortem of the last-known pricing. Browse the pricing blueprint for fully-researched company profiles.

Want to compare MultiOn against living agent-platform pricing? Browse the pricing blueprint for current per-request and per-task agent pricing models.