Pricing

Pricing that reads like a receipt.

Start free with 500 ops a month. Scale to $29, $99, or $299 when your agent does. One flat counter across every endpoint: no credit multipliers and no per-minute metering. The numbers on this page are the numbers in the API and on the receipt.

Founding
$0

Free while in beta. No card. 500 ops a month.

500 ops a month, across every endpointEvery read scored (render_quality)5 MB files · 1-hour retentionNo credit card
Indie
$29/mo

3,000 reads, 20 watchers, 45 conversion routes, $5 extraction credit.

3,000 ops a month20 watchers · 45 conversion routes$5 extraction credit a month · rolls over15 MB files · 7-day retentionOpt-in overage: $0.02 per extra opEmail support
StudioMost teams start here
$99/mo

For teams scaling an agent past the demo.

15,000 ops a month$15 AI credits a month · rolls over100 watchers · 50 MB filesWebhooks + ZIP outputDedicated rendering poolOpt-in overage: $0.02 per extra opPriority support
Production
$299/mo

For when your agent is the product.

50,000 ops a month$40 AI credits a month · rolls over500 watchers · 150 MB files · 30-day retentionDedicated pool · custom rendering engineOpt-in overage: $0.02 per extra opNamed support contact
Every plan includesrender_quality on every readMCP server + REST APISSRF-screened URLsPublic status page + changelogCancel anytime
The unit

One op is one read.

Some meters charge by render time. Some by credit multipliers that change per endpoint. An EnConvert op is one read: a page that takes 30 seconds to render costs the same as one that takes two.

Convert

One op per file or URL across 45 conversion routes, regardless of pages. A batch counts each file.

POST /v1/convert
Perceive

One op per URL read, as Markdown, JSON, or a screenshot. A batch of 20 URLs is 20 ops.

POST /v2/perceive
Lookup

One op per query, plus one for each source page fetched to answer it: web_search with perceive_top=N costs 1+N ops.

POST /v2/lookup
Distill

One op per document, counted only when the summary completes.

POST /v2/distill
Ingest

One op per RAG-ready page produced, not per attempt.

POST /v2/ingest
Discover

One op per call. Map a site before spending ops on the wrong pages.

POST /v2/discover
Watch

Not an op but a slot. Watchers and LLM extraction are paid-plan features: Indie and up include always-on watchers with a webhook or email on every diff.

POST /v2/watch

Blocked reads are not billed: a read that comes back with is_blocked, or whose deductions include http_error or login_wall, never touches your allowance. A completed read that renders badly for other reasons still counts, and comes back with its render_quality and deductions so you know what you paid for. Cache hits count.

Every op draws down the same flat counter, whatever the verb: a convert, a read, a query, a page — each is one op. Watchers are the exception: a persistent allocation you hold, not consumption.

One counter, every endpoint. No multipliers, no buckets.

Compare plans

Every limit, in one place.

Founding$0Indie$29/moStudio$99/moProduction$299/mo
Monthly allowance
Ops per month5003,00015,00050,000
AI credits per month · rolls over$0$5$15$40
Active watchers20100500
Limits
File size limit5 MB15 MB50 MB150 MB
Retention1 hour7 days7 days30 days
Overage$0.02 per op · opt-in$0.02 per op · opt-in$0.02 per op · opt-in
Platform
Webhooks + ZIP output
Dedicated rendering pool
Custom rendering engine
SupportEmailPriorityNamed contact

Ops reset each billing cycle; unspent AI credits roll over. Every value here is enforced by the API: hit a limit and the same number comes back in the 402.

Enterprise

For teams where the agent is the infrastructure: custom volume and SLA, custom AI credits, dedicated infrastructure, SSO and audit logs, invoice billing, and a named account manager. From $1,000 a month.

[email protected] · the founder replies
Add-on

Cloud storage, if you need it.

Every plan cleans up after itself, and converted artifacts expire on the retention schedule. If your outputs should outlive it, add storage: files persist up to your quota for as long as the add-on is active.

Mini5 GB$1/mo
Basic25 GB$3/mo
Standard100 GB$9/mo
Large500 GB$29/mo
Mega1 TB$49/mo

Attach to any paid plan from the dashboard. It’s a separate subscription: cancel it anytime and your files stay until the end of the period you’ve paid for.

Pricing questions

500 ops a month across every endpoint: convert, perceive, lookup, discover, and each new endpoint as it goes live. Every read comes back scored. It’s free while we’re in beta, with no credit card. Artifacts are kept for one hour, long enough to collect your output. Watch and LLM extraction (schema, AI credits) are paid-plan features.
One unit of work. One file converted, one URL perceived, one query looked up, one page ingested, one site-mapping call. A batch of 20 URLs is 20 ops. Everything draws from one flat counter: no credit multipliers, no per-verb buckets, no per-minute metering. A page that takes 30 seconds to render costs the same as one that takes two.
Blocked reads are not billed. When a read comes back with is_blocked true, or its deductions include http_error or login_wall, it never touches your allowance and the response says billed: false. A completed read that renders badly for other reasons, an empty SPA shell or a slow render, still counts, because the render happened; it comes back with its render_quality and deductions so you know what you paid for. Cache hits count as one op.
Because the meters are different. An EnConvert op is flat: no credit multipliers for JSON, screenshots or stealth rendering. Blocked reads are not billed. Files run on the same meter, one op per file regardless of pages, so there is no second conversion vendor. And it is a two-person team that answers email, usually the same day.
The API tells you, honestly: a 402, not a degraded result. On every paid plan — Indie, Studio, and Production — you can opt in to overage at $0.02 per extra op. It’s off by default, so a runaway loop can’t surprise you. On the free Founding tier the counter is a hard limit until the next cycle; watchers stay untouched, because they’re an allocation, not consumption.
Ops don’t: the counter resets at the start of each billing cycle. Unspent AI credits do roll over to the next cycle. If you consistently end the month with ops to spare, you’re one plan too high. Downgrades take effect at the next cycle, and we’d rather you pay for what you use.
Yes. Upgrades, downgrades, and cancellations all happen from the dashboard. Plan changes take effect at the start of the next billing cycle, and after cancelling you keep paid features until the end of the period you’ve paid for.
Probably not. Every plan keeps converted artifacts for its retention window (an hour on Founding, a week on Indie and Studio, 30 days on Production), then deletes them. If your outputs need to outlive that, the storage add-on keeps files up to your quota for as long as it’s active, from 5 GB at $1 to 1 TB at $49 a month.
PayPal: balance, linked bank account, or any major credit or debit card through PayPal. Annual billing and invoicing aren’t self-serve yet; if you need them, email [email protected] and we’ll sort it out.
A fair question, and the one we’d ask. The MCP server and SDK are MIT-licensed, so your integration layer can’t be orphaned. We ship in the open with a public changelog and status page, and pricing changes land there before they land on any invoice. The meter is boring on purpose: one op, one read, one bill.

The meter is honest. So is the free tier.

Start on Founding and find out what your agent actually uses. Upgrade when the 402 tells you to, not before.

Free while in beta. No card. 500 ops a month.