Daytona

The daytonaio/daytona repository (71,722 stars, AGPL-3.0) carries a banner as of this writing: "This repository is no longer maintained. As of June 2026, Daytona's core development has moved to a private codebase. This repository will receive no further updates, fixes, or releases." (https://github.com/daytonaio/daytona). GitHub has not flagged the repo archived: true at the API level, but the commit history backs the banner up: the last commit to main is the maintenance notice itself, dated 25 June 2026, and nothing has landed since (https://api.github.com/repos/daytonaio/daytona/commits). The repo remains public under its original AGPL-3.0 licence (v0.190.0 is the last tagged release, 23 June 2026) and will keep building and running as-is — there just won't be new features, bug fixes, or security patches shipped to it.

What's still open is a separate daytona/clients repository (SDKs, API clients, CLI/MCP tools) that is actively maintained — last push 11 September 2026 — and dual-licensed: the CLI under AGPL-3.0, everything else under Apache-2.0 (https://github.com/daytona/clients, https://raw.githubusercontent.com/daytona/clients/main/LICENSE). So a buyer who came to Daytona expecting an open-core product they could self-host and inspect end-to-end now gets open client libraries talking to a closed server. The commercial product hasn't slowed down — the changelog shows a release roughly every 1-2 weeks through September 2026 (https://www.daytona.io/changelog) — but that development now happens outside public view. There is no self-hosting documentation left on the current docs site.

Daytona's pitch is sub-90ms sandbox creation with full isolation (dedicated kernel, filesystem, network stack) for running untrusted AI-generated code — a claim the vendor makes and we have not independently benchmarked, but it's specific rather than vague marketing (https://github.com/daytona). What is corroborated: several substantive, named customer case studies rather than logo-wall claims. WRITER's CTO is quoted saying Daytona now provisions roughly 300,000 sandboxes a month for their enterprise agent platform, citing 90% infrastructure-cost savings versus their prior in-house AWS Fargate build (https://www.daytona.io/customers/writer, vendor-published case study). Stanford's CS department, SambaNova, and Snorkel are also named customers with their own case studies (https://www.daytona.io/customers) — these are still vendor-published and should be read as testimonials, not audited results, but named enterprise and academic customers willing to be quoted is a stronger signal than most listings on this site can offer.

Billing is genuinely usage-based and per-second: compute at $0.0504/vCPU-hour, memory at $0.0162/GiB-hour, storage at $0.000108/GiB-hour after the first 5GB free, plus a GPU menu spanning RTX 4090 ($0.57/h) up to B200 ($3.59/h) and MI355X ($3.44/h) (https://www.daytona.io/pricing, verified 2026-09-10 per the page's own timestamp). New accounts get $200 in free compute; a startup program offers up to $50k in credits.

Daytona runs a public SafeBase trust center (https://trust.daytona.io) that goes further than most vendors in this index: it discloses named subprocessors (AWS, OVHcloud, Latitude.sh, Lago, Stripe), a SOC 2 Type I report, and a SOC 2 Type II report the vendor says was issued 28 July 2026 by Insight Assurance covering 16 December 2025 through 15 March 2026, plus a HIPAA attestation. It also publishes a named CVE disclosure: CVE-2026-31431 ("Copy Fail"), a Linux kernel AF_ALG/algif_aead vulnerability disclosed 29 April 2026 that could theoretically let one sandbox tenant corrupt kernel page-cache content visible to a co-tenant sandbox on multi-tenant hosts. Daytona states it remediated across all regions within 12 hours, found no evidence of exploitation, and confirms its Sysbox container-isolation boundary was not breached (https://trust.daytona.io, vendor security-advisory update). Publishing a CVE against your own infrastructure, with a specific remediation timeline, is unusual disclosure and worth noting as a positive signal — though it is self-reported and we have no independent confirmation of the 12-hour figure.

Daytona (legally Daytona Platforms Inc.) was founded in 2023 by a team with prior experience building Codeanywhere. Disclosed funding is a $2M pre-seed round (announced November 2023, per PR Newswire and TechCrunch coverage) and a $5M seed round (announced June 2024, led by Upfront Ventures per Dot LA's newsletter coverage) — roughly $7M total across the two rounds (https://www.daytona.io/press). We found no publicly reported funding round since June 2024; that is an absence in what we could find, not a claim that none occurred. The product pivoted from a self-hosted "development environment manager" (its original open-source positioning through 2024) to AI-sandbox infrastructure, which is the product described throughout this listing.