On the blog, we’ve just shared an update on estimated Forge pricing taking effect Jan 1, 2026. Prices are indicative only, and finalized pricing will be shared 6 months before it takes effect.
To summarize, Forge will remain free through 31 December 2025. From 1 January 2026, Forge will follow a consumption-based pricing model, with a generous free tier. Upon exceeding this free threshold, developers will be charged monthly in arrears for consumption of services.
The 100,000 GB-seconds refers to the duration of compute multiplied by the memory (in GB) used by Forge compute. Forge compute uses 0.5GB, so essentially this would be your compute duration in seconds multiplied by 0.5.
For Forge KVS storage, the free threshold of 0.1GB applies per app (across all users of the app). The pricing mentioned is for each additional GB (not 100MB).
My guess for many businesses: Probably fine for many balance sheets compared to other costs. Although apps that have heavy data processing will notice it or will have to use remotes.
The developer in me: I am looking around at cloud pricing, which has usually a high profit margin, and this seems roughly a 10x price of a cloud platform. Probably more 1.5 -2x see update below.
For that price you get a worse Dev platform: Less options, worse tooling, worse diagnostics, heavy restrictions. You need extra Dev effort to deal with the restrictions.
It is not ‘pay more and have way less effort’: It is more ‘pay to play’ + effort, so you can get a ‘runs on Atlassian’ badge =(.
Ok, my summary:
My expectation/hopes: Close to AWS pricing. As Atlassian wants vendors on their platform. Taking the benefit of more apps & Marketplace revenue.
Reality: Quite a extra margin on top of AWS pricing.
** Update: Coud pricing is complicated. That 10x number was me guessing roughly based on raw AWS Dynamo pricing, but I probably go it wrong. In the AWS Dynamo examples 3 millions read/writes are about 2.60. In forge it is by data, size, but assuming 0.5KByte per read/write => 0.5k * 3million => 1.5 GByte: 1.5*2.1USD = 3.15USd. Roughly 1.5x the price.
The ‘function’ pricing seems about 1.4x from the AWS price.
what is the underlined operational service used for Forge Logs - Data Written and Forge KVS: Data Stored? Dynamo, S3 or something else?
For KVS Data Read / Write : Atlassian pricing is only based on the data volume that is transitioning (inbound/outbound) which is quite unusual ; for example, AWS is charging per number of request units ( DynamoDB ). Could you also confirm what is the underline service(s) used and if this pricing will change to the number of requests (reads/write) units in the future?
Thanks
When forge app module are loading (for example jira:projectPage) - each time executing the resolver function resolverfunc (see code below ). The question is - will we pay for the numbers of the function calls when some of limits run out?
Usage is measured across all environments (dev + staging + production)
Your understanding is correct, the amount is measured per month and will reset to zero every new month. The free threshold is designed in a way that most apps will not see charges until they hit a certain scale.
I will confirm and get back to you on whether deployments count towards compute usage.
Thank you for the post, this is quite interesting.
I’m wondering about cases where a customer lets a license expire, but does not uninstall the app. They could still produce costs for us (for example by data storage), but would stop producing revenue. Will there be a way for the owner of an app to remove that app from a customer’s instance or a similar mitigation?
We use various services to run each Forge capability, covering functionality like shard management, data residency and compliance processes, observability, backup and disaster recovery. Forge prices aren’t a one-to-one mapping to a single underlying service.
The prices for KVS are for the data written and read via the Forge APIs, not the data transitioning in and out of the Forge platform. We currently do not have any intention of charging separately for the number of read/write requests.
Thanks for sharing your feedback. This an issue that we are aware of and will resolve this year before we start charging for Forge. I’ve created a public ticket to track this as well, you can follow along for details - Jira
The Forge deployment process does not count as usage. However, if your app subscribes to the event avi:forge:upgraded:app and runs some process when customers upgrade your app, then you would be charged for the execution time of that function.
You would pay for the execution time of the resolver function, measured in GB-seconds, once the free threshold has been passed. Forge Functions are configured with 0.5 GB of memory, so each second of execution time is 0.5 units.
You will not be charged for the number of function calls, only the execution time.
Could you please try to provide more details on how that computation time works and is counted?
I assume that each application will have its own container, but then multiple such containers will be collocated on a physical box. If one of the applications is more cpu intensive and delays the execution of another application on the same box, the one delayed will simply pay more?
Will there be any Jira code running on that box that could potentially delay processing of application’s functions? Will there be a legal agreement that will specify these details?
thanks @reddy1 for those clarifications.
I think there is a misunderstanding regarding the value shown in the Developer Console - Usage - Total Storage.
The value displayed under Usage (GB) is the monthly cumulative (sum) of the daily storage usage. It should instead be the average or the value of the last calculated day .
