Sentry Pricing in 2026: What You Actually Pay

Sentry sells itself on a clean sticker: free to start, then $26 a month, then $80. Those numbers are real, and they are almost never what you pay. Sentry's 2026 pricing is event-based, which means the plan fee buys you a fixed bucket of errors, tracing spans, replays and logs, and everything above the bucket is billed by the unit. For any app with real traffic, that overage is the bill. The sticker is a rounding error.

This is the arithmetic behind Sentry's pricing page and its billing documentation, current as of July 2026, with the per-unit rates converted into numbers you can actually reason about, three modeled 30-day bills, and the reason a $26 plan and an $80 plan include the exact same base quotas.

Sentry is an error-monitoring and application-performance platform that bills on the volume of data you send it, not on seats. Five categories meter separately: errors, tracing spans, session replays, logs, and a handful of small line items (attachments, cron monitors, uptime monitors, profiling hours). You buy a plan, the plan includes a baseline of each category, and anything past the baseline draws down a shared pay-as-you-go budget.

The three plans, and the trap hiding in them

Here is what each plan costs and includes on the current 2026 model, billed annually.

Read the Team and Business rows again. They are identical. Same 50,000 errors, same 5 million spans, same 50 replays, same 5 GB of logs. Upgrading from Team to Business does not buy you a single additional included event. It buys features: single sign-on, advanced dashboards, finer-grained quota and spend controls, insights, and longer sampled retention. That is a defensible reason to pay triple, but only if you need the features. If you are on Business because you assumed it came with more headroom, you are paying $54 a month for nothing you use.

The free Developer plan is more generous than most people realize. It includes the full 5 million tracing spans and 5 GB of logs, the same as the paid tiers. The only things it starves are errors (5,000 versus 50,000) and users (one seat). A solo developer can run real tracing and logging for free until error volume crosses 5,000 a month.

The per-unit rates, in numbers you can use

Sentry publishes overage prices as per-event decimals with a lot of zeros, which is how the bill stays mysterious. Converted into readable units, here is what you pay above your included baseline. Rates are tiered, so the low end applies to the first slice of overage and climbs with volume.

Two things jump out. First, session replays and logs cost the same on both plans, but errors and spans do not: Business charges roughly two to three times the Team rate for the same overage event. So the plan that costs more per month also costs more per extra error. If your bill is dominated by overage rather than features, Business can be the more expensive choice, not the cheaper one. Second, spans are priced per million and a traced app emits them in the tens or hundreds of millions. That single line quietly becomes the largest number on the invoice.

Three real 30-day bills

None of these is a quote from Sentry; each is the published rate math applied to a stated traffic profile. Your mileage varies with your own volume, but the shape holds.

Bill 1: solo SaaS on Team ($26). Assume 150,000 errors, 15M spans, 1,500 replays, 8 GB of logs in a month.

The $26 sticker is 42 percent of a $61 bill. Manageable, but already more than double the advertised price.

Bill 2: growth-stage app on Business ($80). Assume 800,000 errors, 120M spans, 25,000 replays, 30 GB of logs.

The $80 plan fee is 11 percent of a $700 bill. Errors and spans together are 80 percent. This is the tier where finance starts asking questions.

Bill 3: high-traffic app on Business. Assume 3M errors, 400M spans, 80,000 replays, 80 GB of logs, 300 profiling hours. With tiered rates climbing at this volume, the errors and spans lines blend upward.

This is the "why is our observability bill more than our cloud bill" thread on Reddit, explained. Nobody chose to spend $2,500 a month. Tracing was switched on, span volume scaled with traffic, and the meter ran.

Why spans, not errors, blow up the bill

Most teams adopt Sentry for errors and think of it as an error tool. Then someone enables performance monitoring or distributed tracing, and every request starts emitting spans, often dozens per request. A moderately busy API can produce hundreds of millions of spans a month without anyone noticing, because spans do not show up in your face the way a spike in errors does. At roughly $2 to $4 per million, 300 million spans is $600 to $1,200 on its own. In all three bills above, spans are either the largest line or a close second.

The good news is that spans are also the most compressible line, because you rarely need all of them. Trace sampling, keeping a representative fraction rather than every trace, is the single highest-leverage move on a Sentry bill. Dropping to 10 percent sampling on a span-heavy app cuts the largest line item by ninety percent while preserving enough traces to debug with.

How to actually cut a Sentry bill

• Sample your traces. This is the biggest lever by a wide margin. Set a trace sample rate well below 100 percent and raise it only when investigating something specific.
• Filter noisy errors at the source. Inbound filters and per-project rate limits stop known-junk errors (bot traffic, a single flapping integration) from eating your error quota before they are ever billed.
• Sample session replays. Full-session replay on every user is expensive and mostly unwatched. Capture replays on error, or a small percentage of sessions, not all of them.
• Reserve volume and pay annually. Annual billing is what unlocks the $26 and $80 stickers in the first place, and pre-committing reserved volume is cheaper per unit than pure pay-as-you-go for predictable baselines.
• Mind the shared budget. Your pay-as-you-go budget is one pool split across all categories, first come first served. When it runs out, Sentry drops the excess data rather than charging you, which protects your wallet but silently blinds you. Set the budget high enough to keep visibility, then cut usage at the source rather than letting data get dropped at random.

When the answer is a different tool

If your bill is climbing and the features are not pulling their weight, look sideways. SigNoz is an open-source, self-hostable observability stack that trades a monthly invoice for the cost of running it yourself, which pencils out once your Sentry bill clears a few hundred dollars a month. At the heavier, more integrated end, Datadog offers broader infrastructure and APM coverage, though its own usage-based pricing has an identical failure mode: the sticker is not the bill. The pattern to internalize is the one from every usage-priced tool, the same one behind the Claude API pricing teardown and the Cloudflare R2 pricing breakdown: the advertised number is the floor, and the meter is where the money goes.

Math check: on the growth-stage bill, the $80 plan fee is 11 percent of the total and tracing spans alone are 48 percent. Budget Sentry by span volume and sample rate, not by the plan sticker.