Author
Trigger.dev pricing in 2026 runs on two meters: $0.25 per 10,000 runs plus a per-compute-second charge set by machine size, drawn from monthly credits (Free $5, Hobby $10, Pro $50). A light side project stays free; a busy agentic workload can pass $1,000/month, 99% of it compute. Because waits are checkpointed and not billed, moving in-process sleeps to native waits is the single biggest lever.
As of July 2026, Knock's notification API charges per message, where a message is one delivery to one user on one channel. The Starter plan is $250 per month for 50,000 messages, then $0.005 each. Because Knock meters per channel, a single notification sent to email, push, and in-app counts as three messages, so a 5,000-user app fanning three channels can run about $750 per month. The same alerts cost roughly $250 on Courier's per-notification meter and about $54 on Novu's per-run Pro plan. Match the meter to your channel fan-out before you pick a vendor.
Axiom bills on two meters, data loaded and query compute, not host count. A 2026 teardown with the free tier, the $25 flat band, three worked bills and the query-compute trap that quietly doubles the invoice.
Langfuse pricing in 2026 starts free (Hobby: 50,000 units per month), then $29/month (Core), $199 (Pro), and $2,499 (Enterprise), with extra units at $8 per 100,000. The catch: a unit is every trace, observation, and score, so one multi-step agent request can burn 20 or more units. That is why a busy agent app pays far more than its request count suggests.
Inngest pricing meters every step, not every job. Here are the real 2026 rates, three worked 30-day bills, and the two meters (step count and concurrency) that actually spike your invoice.
As of July 2026, PostHog has no fixed subscription. It is free to start with no credit card, and each of its products bills separately once you pass a generous monthly free tier: 1M product-analytics events, 5K session replays, 1M feature-flag requests, 100K error-tracking exceptions, and more. Product analytics starts at $0.00005 per anonymous event and gets cheaper per event as volume rises. The catch nobody flags: identified events can cost up to 4x more. Most side projects pay $0; a growing SaaS usually lands between roughly $50 and a few hundred dollars a month; the five-figure horror stories come from high-volume, autocapture-heavy, all-identified setups with no billing limit set.
As of July 2026, Upstash Redis is free up to 256 MB and 500K commands per month; pay-as-you-go bills $0.20 per 100K commands plus $0.25/GB stored, and fixed plans run $10 to $1,500/month. Because Upstash charges per command, not per server, request volume sets your bill, not data size. Above ~5M commands/month a flat $10 fixed plan usually beats pay-as-you-go.
Pinecone runs on usage-based serverless pricing in 2026: a free Starter tier, a $20/month flat Builder plan, and a Standard plan with a $50/month minimum. You pay $16 per million read units, $4 per million write units, and $0.33 per GB of storage each month. The catch most teams miss: a query costs one read unit per GB of namespace size, not per query, so your bill is set by how you shape your namespaces, not how many searches you run. A hobby RAG app fits the free tier at $0; a 20 GB single-corpus production app runs about $110/month; a 50-million-vector multi-tenant app with per-tenant namespaces lands near $75.
Cloudflare Workers bills CPU time, not duration, which inverts everything you know from AWS Lambda. Here is the 2026 pricing, three worked monthly bills, and the 3-millisecond breakpoint that decides your cost.
As of July 2026, Neon uses usage-based pricing with no flat subscription: a Free tier ($0), Launch ($0.106 per CU-hour) and Scale ($0.222 per CU-hour). Because Neon bills compute by the CU-hour and scales to zero when idle, the same 1 CU Postgres database costs roughly $0 to $16 a month if it sleeps, about $81 a month always-on on Launch, and about $162 always-on on Scale. Your bill is set by uptime, not by data size.
Render's advertised prices hide the real bill. In 2026 your cost is three stacked layers: a workspace plan, per-service compute, and metered usage. A solo production app starts near $13/mo (not $7), a small team app around $115/mo. Here is the arithmetic, with three worked 30-day bills.
Sentry pricing in 2026 is event-based: a free Developer plan, Team at $26/mo and Business at $80/mo (billed annually), each including 50,000 errors, 5M tracing spans, 50 replays and 5 GB of logs. Everything above that baseline is pay-as-you-go, and that overage, not the sticker, is your real bill. Errors run about $0.15 to $1.11 per 1,000 and spans about $1.60 to $4.00 per million, so tracing usually dominates. Below: the full per-unit ladder, three modeled 30-day bills, and why the Business plan includes the same base quotas as Team at three times the price.
