Korve vs InsForge
Korve vs InsForge: separate resources or one bundled backend?
InsForge packs a full backend — Postgres, auth, storage, functions, sites — into one server per project, with a free tier and per-user pricing. Korve treats the app and each of its data resources as independently operated primitives with pure usage pricing. Both are built for coding agents; the difference is whether the database is part of the app server or a separately sized, separately recovered resource.
Korve is a paid-only cloud for agent-built applications. A coding agent (or a human) deploys a repository to a stable *.korve.app URL and provisions only the managed resources the app declares — Postgres, object storage, caches, queues, cron jobs, budgets, and custom domains — through one typed operation spec shared by the API, dashboard, native CLI, MCP server, and manifest. Pricing is pure usage with a $5/month minimum and no free tier.
Primary-source review updated July 27, 2026, including a logged-in hands-on session. Verify current vendor terms before buying.
Where the boundaries actually differ
Both platforms give an agent a CLI, an MCP server, and skills. The differences are in how the infrastructure underneath is sized, priced, recovered, and exposed.
| Dimension | InsForge | Korve |
|---|---|---|
| Database | Postgres runs inside the same per-project backend server as the API. Growing the database means resizing the whole box; vertical scaling is the documented scale path. | A separate resource with its own SKU, region, and disk. Size the database for the data and the runtime for the traffic — each moves independently. |
| Recovery | Manual snapshots with in-place restore, plus daily backups on paid plans, and no published high-availability topology or automated-failover contract. Its data processing agreement additionally claims encrypted point-in-time recovery and multi-availability-zone backup storage; the product surface and pricing page do not expose a customer-selectable recovery window, so confirm what is actually available to you. | Every database SKU includes 14-day point-in-time recovery plus 30 daily, 12 weekly, and 12 monthly backups. High availability adds two replicas, and resizes roll node by node with no maintenance window. |
| Idle behavior | Free projects pause after one week of inactivity; paid projects run an always-on backend instance that does not pause. | Paid-only, so there is no pausing tier at all. Scale-to-zero runtime compute idles to $0 and cold-starts on request; declared stateful resources keep their published meters, and always-on apps never idle. |
| Pricing model | $25/month Pro with 100,000 monthly active users included, then $0.00325 per additional monthly active user — the bill tracks how many people use the app each month as well as what it consumes. | Pure usage against a public rate card: compute, database, storage, egress. Monthly active users are not a meter. $5/month minimum, credited against usage. |
| Endpoints | Assembled from many third-party clouds that are named across its docs and surfaces. One upstream runtime migration required customers to redeploy before the provider's shutdown date. | One published set of endpoints — *.korve.app and *.korve.dev — with the substrate underneath treated as an internal implementation detail, so a provider change is designed to be a platform migration rather than a customer redeploy. Korve has not yet had to prove that in public; InsForge names its upstreams, which is more transparent and also means an upstream lifecycle event can reach the customer. |
| Agent surface | JavaScript CLI installed through npx, remote OAuth MCP, agent skills, and a reviewable config plan/apply loop. | One typed operation spec shared by the API, dashboard, native Rust CLI, MCP, and manifest — every surface exposes the same contract. |
What one modeled configuration costs on each rate card
Take InsForge's own representative Pro application — one always-on backend, a 10 GB database, 300 GB of bandwidth, 150 GB of files, a million function calls — and it bills roughly $69.80/month at or below the 100,000 included monthly active users. A roughly analogous set of Korve line items — an always-on shared instance, a 10 GB managed database, the same egress and object storage, plus logs — comes to $77.81/month. At that size Korve is the more expensive of the two.
What changes is the slope. Every additional monthly active user past 100,000 costs $0.00325 on InsForge, so the same app at 150,000 monthly active users adds $162.50/month for the extra 50,000 users alone, before any infrastructure moves. On Korve there is no user meter: an app with 150,000 users and an app with 100 users read the same rate card and pay for compute, database, storage, and egress. On these exact modeled line items, the raw bills first meet at about 102,465 monthly active users.
Korve's minimum is $5/month, credited against usage, and there is no free tier to start from.
InsForge figures are its published Pro allowances and overage rates, reconstructed in our primary-source review; Korve figures are its published rate card. The products do not package equivalent CPU, memory, database, or edge-function capacity, so this is not a performance-normalized crossover and a workload that needs a larger Korve runtime will cross later or not at all. Neither figure is a quote — benchmark your app and check both pricing pages before modeling a bill.
A database that scales without touching the app
InsForge runs each project's Postgres inside the project's backend server, so the database and the API scale as one unit, and the documented growth path is a bigger box. Its multi-region announcement lists cross-region replication and automated failover as future work.
Korve provisions databases as their own resources: pick a SKU and region per database, add high availability with two replicas when the app warrants it, and resize a high-availability database node by node while the app keeps serving — replicas restart one at a time and the primary switches to an already-resized replica. Point-in-time recovery over a 14-day window is part of every database SKU, not an add-on.
Korve does not claim whole-backend branch/merge parity with InsForge; database branches cover the database, not the full project.
Where InsForge fits
Pick InsForge when any of these match your situation.
- You want a free tier: InsForge lets you try the whole workflow — database, auth, storage, functions, deploys — without a card, which Korve deliberately does not offer.
- You want whole-backend branches: InsForge copies the entire backend (schema, auth, buckets, functions, schedules) and merges it back with a three-way diff.
- You want an open-source escape hatch: the core platform is Apache-2.0 and self-hostable with a compose file.
- You are shipping mobile: first-party TypeScript, Swift, and Kotlin SDKs cover more than web-only teams.
- You want a broader backend component catalog: InsForge includes transactional email, vector search, edge functions, and whole-backend branches. Korve focuses on managed application infrastructure and operations rather than matching every component.
- Your app is under 100,000 monthly active users and resource-light: on the representative workload above, InsForge's bill is the smaller one.
Choose Korve when the database deserves its own sizing, recovery, and availability story, when the bill should track resource usage rather than user count, and when you would rather not re-point clients at a new upstream URL. One caveat in the other direction: a single-node Korve database restarts to resize, and the node-by-node path requires high availability, which starts above the single-node SKUs.