Diffbot pricing

About

Diffbot builds and sells structured web data. Its core products are a suite of web-extraction APIs — Extract (turn any page into structured JSON), Crawl (spider a site for pages to extract), and Natural Language (entity and relationship extraction from text) — plus the Diffbot Knowledge Graph (DKG), which the company describes as the world’s largest autonomously-built knowledge graph of organizations, people, articles, places, and things. The same APIs Diffbot uses to construct the DKG are published for customers to build their own data pipelines.

The company sells to developers and data teams who need web data at scale — market intelligence, news monitoring, machine-learning training sets, and e-commerce product data are its headline solutions. Customers range from individual researchers on the Free plan up to large enterprises; Diffbot has cited users such as DuckDuckGo, Bing, Snapchat, and Adidas. Founded by CEO Mike Tung (the company traces its roots to 2008–2012), Diffbot is privately held and headquartered in Menlo Park, California. It raised a $10M Series A in 2016 led by Tencent and Felicis Ventures and has otherwise grown without large outside capital — Tung describes Diffbot as one of the few bootstrapped, profitable AI companies, reportedly reaching roughly $3.1M revenue in 2024 with a ~35-person team.

Diffbot’s strategic bet is contrarian in the LLM era. Rather than cramming knowledge into model weights, it grounds AI in a structured, continually-refreshed graph of “verifiable facts” — Tung’s thesis is that “you don’t actually want the knowledge in the model; you want the model to be good at using tools so it can query knowledge externally.” Diffbot’s own knowledge-grounded LLM has been reported to beat Gemini and ChatGPT on the real-time-factuality FreshQA benchmark (81%). That positioning matters for pricing: the Knowledge Graph is both the product and the moat, which is why graph operations are the most credit-expensive activities on the meter.

Pricing is positioned as low-friction and transparent: plans start at Free with no credit card and no contract, and the published tiers disclose exact monthly fees, credit allotments, per-credit rates, and rate limits — only the Enterprise tier is quote-based.

Pricing summary : How Diffbot’s prepaid credit model and per-credit overage work

Diffbot uses a hybrid credit model with two dimensions:

1. Flat monthly plan fee: Each paid tier carries a fixed monthly price — Startup $299/mo, Plus $899/mo, Enterprise custom — that bundles a prepaid, monthly-renewing credit allotment (250,000 / 1,000,000 / custom credits respectively). The Free plan is $0/mo with 10,000 credits.
2. Per-credit overage: Once the bundled allotment is consumed, additional usage is billed pro rata at the plan’s per-credit rate — $0.001/credit on Startup, $0.0009/credit on Plus. The per-credit rate falls as you climb tiers, so higher plans double as a volume discount.

Credits are spent at activity-specific rates: extracting one web page is 1 credit, exporting one Knowledge Graph entity record is 25 credits, and a summarized facet query is 100 credits. Plans also differ on rate limits (5 calls/min on Free; 5, 25, and 25+ calls/sec on the paid tiers) and on product access — Crawl is gated to Plus and above.

What makes this different: the same credit pool meters two very different cost structures — cheap per-page web extraction (1 credit) and expensive Knowledge Graph record exports (25–100 credits) — so a customer’s effective unit cost depends entirely on which Diffbot product they lean on.

Pricing by product

Diffbot platform (self-serve plans)

Sales motions across products: self-serve / PLG for Free, Startup, and Plus; sales-led (Schedule a Demo) for Enterprise.

Credits per processing activity

Credits are consumed at different rates depending on the operation. The published rate table:

Note: Crawl itself consumes 0 credits to spider — the credits are charged on the Extract operations it triggers. Each plan is also subject to API rate limits (calls/sec), with a fuller table in Diffbot’s docs.

Diffbot for Students (academic program)

Active students in college/university (or higher) aged 16+ can apply for free access to the entire Diffbot suite — Extract, Crawl, Knowledge Graph, and Natural Language — at no charge. The program is application-gated (a ~5-minute form) and also grants project-listing and community-forum access. Personal, commercial, non-profit, and research projects are all eligible.

Hidden costs : What Knowledge Graph-heavy and high-volume crawling teams actually pay

Diffbot’s headline prices look cheap, but the bill is driven by which activity you run, not which plan you pick. Because one credit buys 1 page extract but only 1/25th of a Knowledge Graph entity export, two teams on the same plan can pay wildly different effective rates. Two archetypes make this concrete.

Archetype A — Knowledge Graph-heavy lead-gen team on Plus. A sales-intelligence team on Plus ($899/mo, 1,000,000 credits @ $0.0009 overage) exports 50,000 enriched company/person records per month. At 25 credits per entity export, that is 1,250,000 credits — more than the entire Plus allotment.

The lesson: a Knowledge Graph workload exhausts the allotment ~25× faster than the same number of page extracts, so the “1,000,000 credits” headline translates to only ~40,000 entity records. A team modelling its bill from page-extract intuition will under-forecast badly. (See our guide to choosing the right usage metric for why a single meter spanning two cost structures is hard to reason about.)

Archetype B — high-volume crawler on Plus. A market-intelligence team uses Crawl (0 credits to spider) to feed 2,000,000 page extracts/mo, all on Plus. Crawl itself is free; the Extract operations it triggers are billed at 1 credit each.

If those extracts route through Datacenter Proxy (2 credits each), the same volume doubles to 4,000,000 credits and the overage line jumps to ~$2,700 on top of the plan fee — proxy choice silently doubles the marginal cost. This is the same “the real bill lives in the overage, not the sticker” dynamic we cover in how finance teams should plan for usage-based pricing.

Want to estimate your own Diffbot bill? Use the Diffbot pricing calculator to model your monthly cost based on page-extract volume, Knowledge Graph record exports, and your plan’s per-credit overage rate.

Pricing evolution : From per-call API metering toward a unified credit pool

Diffbot’s pricing is unusually stable for a company this old. The $299 (Startup) and $899 (Plus) anchor prices have held since 2016 — a decade. What has changed over thirteen years of archived snapshots is not the price but the meter (API calls → credits), the product surface (Global Index → Knowledge Graph → LeadGraph), and the free tier (paid trial → free-forever).

Cadence

Tracked range: 2013-12–2026-06. Quarters not listed were verified stable (0 price changes, 0 SKU additions) across the monthly Wayback snapshots. Note the credit migration is dated to the 2019-03 → 2020-04 snapshot gap; Diffbot took no monthly archive between those dates, so the exact migration month is recorded as unknown.

Notable changes

• 2016-03 — Plus $899/mo introduced; Pro repriced from $4,999 to $3,999; the $299/$899 anchors that still define the 2026 page first appear (Wayback snapshot 2016-03-14).
• 2018-04 — “Global Index Calls” added as a separate allotment alongside API Calls, foreshadowing the Knowledge Graph product (Wayback snapshot 2018-04-19).
• 2020-04 — Unit switched from API Calls to activity-weighted Credits; “Global Index” rebranded “Knowledge Graph™”; the Professional $3,999 tier was retired (Wayback snapshot 2020-04-08).
• 2024-04 — Permanent Free plan (10,000 credits/mo, no card) replaced the 14-day trial; page headline changed from “Plans starting at $299/mo” to “Plans starting at Free” (pinned between Wayback snapshots 2024-03-22 and 2024-04-25).

The calls-to-credits migration in detail

The single most consequential pricing change in Diffbot’s history was not a number — it was the meter. Through 2019, Diffbot billed one flat unit: the API call (one extracted page = one call), with a second “Global Index Calls” bucket for graph lookups. Sometime in the 2019-03 → 2020-04 archive gap, Diffbot collapsed both buckets into a single credit pool and made credits activity-weighted: 1 credit per page extract, 2 with a datacenter proxy, 25 to download a Knowledge Graph entity, and 100 for a refresh or a summarized facet query.

This is a textbook example of moving from a SKU-based meter to a unified credit model: instead of maintaining separate price lists for “extraction” and “graph” products, one credit balance now spans both, and the credit cost of each operation encodes its underlying cost to serve. The trade-off is legibility — a customer can no longer read a single number off the page and know what 1,000,000 credits buys, because the answer is “1,000,000 pages, or 40,000 graph records, or 10,000 refreshes.” We unpack the broader industry shift toward this packaging in from entitlements to credits.

What’s unique : Activity-weighted credits across extraction and Knowledge Graph

1. One credit pool meters wildly different unit costs. The same credit is worth 1 page extract or 1/25th of a Knowledge Graph entity export — a 25× spread, rising to 100× for facet queries and refreshes. This is credit-based billing used as a cost-allocation tool: the credit weight of each operation encodes its cost to serve, so Diffbot can price cheap scraping and expensive graph exports off one balance without separate SKUs. The flip side is that a customer’s effective price per “thing” depends entirely on product mix.

2. The meter is the moat-signal. The most expensive operations on the credit table are precisely the ones that touch the Knowledge Graph (25–100 credits) — the asset Diffbot spent fifteen years building. Cheap, commoditised page extraction is 1 credit; proprietary, hard-to-replicate graph access is 25–100×. The pricing literally encodes where the defensibility lives — a sharper answer to the value-metric problem in AI pricing than most AI companies manage.

3. Overage is pro rata, not punitive. Exceeding the monthly allotment simply bills additional credits at the plan rate, rather than throttling, erroring, or forcing an upgrade. This is the prepaid-allotment-plus-overage shape of a hybrid pricing model: a predictable platform floor with a smooth, metered ceiling.

4. The per-credit rate is the real upgrade incentive. Climbing from Startup ($0.001) to Plus ($0.0009) buys both a 4× larger allotment and a cheaper marginal credit — folding a volume discount directly into the tier ladder rather than negotiating it. The rate keeps falling into Enterprise (historically as low as $0.00085).

5. Crawl is free to run, billed on what it triggers. Spidering a site with Crawl costs 0 credits; the Extract operations the crawl fires are what consume the meter. This decouples discovery from extraction in the bill and is why Crawl can be gated as a Plus feature without itself being a metered line item.

Strengths & weaknesses

Billing UX : Credit-metered dashboard, token management, and call-rate limits

• Credit allotment + per-credit overage — every plan shows its monthly renewing credit allotment and the exact $/credit rate (e.g. “250,000 Credits ($0.001 per Credit)”); overage is billed pro rata at that rate.
• Dashboard Access — all tiers (including Free) get the Diffbot dashboard to monitor usage against the credit allotment.
• Token Management — all tiers include API token management for issuing and rotating keys.
• API rate limits (calls/sec) — enforced per plan: 5 calls/minute on Free, then 5, 25, and 25+ calls/second on Startup, Plus, and Enterprise.
• Multiple User Licenses — seat allocation is tier-gated: Free and Startup list multiple user licenses, Plus includes 3, Enterprise is custom.
• No-contract self-serve — paid plans require no long-term contract; Free requires no credit card.

Strategic wins : Why Diffbot’s credit packaging works

1. A free-forever tier lowers developer-adoption friction

10,000 monthly credits with no credit card lets developers prototype a full extraction pipeline before paying, seeding bottom-up adoption ahead of any sales conversation. The 2024 switch from a 14-day trial to a permanent free plan was a deliberate PLG move — a standing free tier keeps Diffbot in the consideration set long after a trial would have expired. This is the freemium pricing playbook executed cleanly: full API access on the free tier, capped only by volume and rate, so the product sells itself before sales does.

2. Activity-weighted credits price by cost, not by SKU

Charging 1 credit per page but 25–100 per Knowledge Graph record lets Diffbot recover the real cost difference between cheap scraping and expensive graph exports without maintaining separate price lists. It is a single value metric that flexes across very different operations — the kind of design our choosing the right usage metric guide argues is the hardest and most important pricing decision a usage-based company makes.

3. Pro-rata overage removes the upgrade cliff

Because exceeding the allotment simply bills more credits at the plan rate, customers never hit a hard wall — usage can spike without a forced plan change or service interruption. Combined with a generous prepaid allotment, this gives buyers a predictable floor and a smooth ceiling, the exact shape we recommend in understanding prepaid credit models.

4. A decade of price stability builds trust

The $299 and $899 anchors have not moved since 2016 even as the meter, products, and free tier were overhauled around them. For a buyer doing multi-year budget planning, that stability is a feature — Diffbot changes packaging, not the sticker, which is a quiet but real differentiator versus AI-era peers who reprice every quarter.

Areas to improve : Closing the Startup-to-Plus packaging gap

1. The $299 → $899 jump needs a middle rung

A team that outgrows Startup’s 250,000 credits but doesn’t need Crawl faces a 3× price jump with nothing in between. A mid tier (~$499, ~500K credits) or an à-la-carte Crawl add-on for Startup would smooth the upgrade path and capture customers who currently either over-buy into Plus or churn at the cliff. Diffbot once ran a five-tier ladder (2016–2019); reintroducing a rung between Startup and Plus would echo that without the abandoned $3,999 Professional tier.

2. Knowledge Graph exports deserve clearer cost framing

At 25–100 credits per record, a modest KG export workload can silently exhaust a plan’s allotment — as Archetype A above shows, 50,000 entity exports blows past the entire Plus allotment. An in-dashboard credit estimator that translates “export 10,000 entities” into a dollar figure, plus proactive threshold alerting before the allotment is consumed, would turn overage from a surprise into a decision.

3. No annual commitment option for predictable buyers

The self-serve plans are month-to-month only; offering a discounted annual prepay would reward committed customers and improve revenue visibility for both sides. For a profitable, bootstrapped company, committed annual revenue is especially valuable — and prepaid annual credit pools fit naturally onto the existing meter, as covered in revenue recognition for usage-based pricing.

Key takeaways

1. Activity-weighted credits let one meter price very different operations. Diffbot charges 1 credit per page but 25–100 per Knowledge Graph record, capturing cost-to-serve differences without separate SKUs. If your product spans cheap and expensive operations, weight the credits — don’t ship two price lists.
2. Bundling a prepaid allotment into the plan fee makes bills predictable. Customers know their baseline before any overage, and the plan fee + included credits reads as a clean floor. Pro-rata overage above the floor avoids the hard-wall churn that hard caps cause.
3. Stable sticker prices buy trust; change the packaging, not the number. Diffbot held $299/$899 for a decade while overhauling the meter, products, and free tier around them. Buyers reward price stability even when everything else evolves.
4. Let the meter encode your moat. Diffbot’s most credit-expensive operations are exactly the ones that touch its hardest-to-replicate asset (the Knowledge Graph). Pricing that mirrors defensibility both monetises the moat and signals where the value is.
5. A free-forever tier beats a trial for PLG. Migrating from a 14-day trial to a standing free plan keeps you in the consideration set indefinitely; the cost of 10,000 free credits is a marketing line item, not a giveaway.

UBP implications

1. Credit pools can unify heterogeneous product lines under one meter. Diffbot bills extraction APIs, Crawl, Natural Language, and a Knowledge Graph from one credit balance, so adding a product means adding a credit weight, not a new price page. The cost is interpretability: the more operations a single credit spans, the harder it is for a buyer to translate “1M credits” into capacity.
2. Per-credit rate as a tier lever folds volume discounting into the ladder. Cheaper marginal credits at higher tiers ($0.001 → $0.0009 → $0.00085) reward growth automatically, replacing a custom volume negotiation with a published rate card. This makes the upgrade math self-evident to the buyer.
3. Pricing migrations can change the meter without changing the price. Diffbot’s calls→credits switch preserved the $299/$899 anchors while completely re-basing what those dollars buy — proof that a unit migration and a price change are separable decisions, and that decoupling them protects buyer trust during a re-platform.

Sources

• Diffbot pricing page (accessed 2026-06-04)
• Diffbot Schedule a Demo (Enterprise) (accessed 2026-06-04)
• Diffbot for Students (accessed 2026-06-04)

Bottom line

Diffbot prices structured web data on a clean, transparent credit model: a flat monthly fee bundles a prepaid credit allotment, and overage is billed pro rata at a per-credit rate that drops as you climb tiers. The clever bit is activity-weighted credits — 1 per page extract, 25–100 per Knowledge Graph record — which lets a single meter span cheap scraping and expensive graph exports.

Want to compare Diffbot against other web-data and scraping pricing? Browse the pricing blueprint.

Pricing timeline : Major events on a vertical axis

Each milestone below corresponds to a public pricing change, product launch, or material adjustment. Major events use a filled marker; minor adjustments use a faded one.

Current snapshot — four-tier credit model

Jun 2026

Free ($0, 10K credits), Startup ($299/mo, 250K credits @ $0.001), Plus ($899/mo, 1M credits @ $0.0009), and custom Enterprise. Crawl gated to Plus+. Overage billed pro rata at the plan's per-credit rate. $299/$899 anchors unchanged since 2016.

Free-forever plan replaces the 14-day trial

Apr 2024

The 14-day trial became a permanent Free plan (10,000 credits/mo, $0/credit, no credit card). The page headline shifted from 'Plans starting at $299/mo' to 'Plans starting at Free.' Crawl access remained gated to Plus and above; LeadGraph appeared in the product nav. Pinned between the 2024-03-22 ($299 trial) and 2024-04-25 (Free) Wayback snapshots.

Migration from API Calls to Credits + Knowledge Graph rebrand

Apr 2020

Between the 2019-03 and 2020-04 snapshots Diffbot replaced 'API Calls' with a unified Credit meter and rebranded 'Global Index' as the Knowledge Graph™. New activity-weighted credit table: 1 credit per page extract, 2 per proxy extract, 25 per entity downloaded, 100 per entity refresh / facet query. The Professional $3,999 tier was dropped, collapsing the ladder back to Free Trial / Startup $299 / Plus $899 / Enterprise. Verified via Wayback snapshot 2020-04-08.

Global Index Calls added as a second metered dimension

Apr 2018

Plans gained a separate 'Global Index Calls' allotment (0 on Startup, 250K on Plus, 1M on Pro) alongside API Calls — the precursor to the Knowledge Graph product. Overage quoted per 1,000 calls ($1 / $0.90 / $0.80). Tiers and headline prices unchanged ($299 / $899 / $3,999). Verified via Wayback snapshot 2018-04-19.

Five-tier ladder — $899 Plus introduced

Mar 2016

Diffbot expanded to a five-tier ladder: 14-Day Trial, Startup $299/mo (250K calls, 5 calls/sec), Plus $899/mo (1M calls, 25 calls/sec), Professional $3,999/mo (5M calls, 50 calls/sec, $.0008/extra), and custom Enterprise. The $299 and $899 anchor prices that persist into 2026 were both in place by this snapshot. Verified via Wayback snapshot 2016-03-14.

Per-call API metering — Standard $299, Pro $4,999

Dec 2013

Earliest archived tier set: 30-Day Free Trial (10,000 included calls), Standard $299/mo (250,000 monthly calls, $.001/extra call), Pro $4,999/mo (5,000,000 calls, $.0009/extra), and custom Enterprise. Billing unit was the API call (one extracted page = one call); no Knowledge Graph yet. Verified via Wayback snapshot 2013-12-20.