Namespace

Managed compute for CI: GitHub Actions runners, Docker builders, Bazel caching and dev boxes on its own hardware.

ProprietaryFreemium#ci#github-actions#runners#docker#bazel#macos

What is Namespace?

Namespace sells build and test compute that runs on its own hardware. For GitHub Actions you connect your organisation and point runs-on at a runner profile or a label such as nscloud-ubuntu-24.04-amd64-4x8; Linux (x64 and ARM), Windows and macOS on Apple silicon are offered, with cache volumes that keep dependencies between runs. The same platform provides remote Docker builders, Bazel remote cache and execution, and Devboxes, cloud dev environments for people and coding agents. The nsc CLI is open source (Apache-2.0).

Namespace Labs builds its own racks, hypervisor and scheduler. The service is proprietary and sold as plans that include unit minutes (1 vCPU and 2 GB for a minute, times 2 for Windows and 10 for macOS). On 5 October 2026 it announced a $42M Series B led by Scale Venture Partners, seven months after an NEA-led Series A, for $65M in total.

This autumn its releases have centred on Devboxes: Windows Devboxes on 9 November and computer use on 11 November 2026. Against Blacksmith, it covers more of the build beyond Actions, but caps concurrency by plan.

Our review

3.5

A strong choice when CI is more than GitHub Actions jobs: macOS at scale, Bazel remote execution and remote Docker builders all run on the same platform with one bill. The plans cap concurrent vCPUs, though, so the queueing you leave GitHub to escape can come back, and its status page logged 16 queue-time incidents between August and mid-November 2026.

Pros

+ Runner configuration can live in the runs-on label, so it is versioned with the workflow
+ One platform for Actions runners, Docker builders, Bazel cache and remote execution, and dev boxes
+ Included unit minutes make a steady Linux workload cheap: 80,000 unit minutes fit in the $100 Team plan

Cons

− Concurrency is capped per plan (64 Linux vCPU on Team), so wide test sharding queues
− Frequent queue-time incidents on its status page in autumn 2026
− Billing in unit minutes, with RAM counted, takes arithmetic to predict

Reviewed by SeekDeploy Editorial · updated 2026-11-15. Reviews are independent — we don’t accept payment for coverage or ratings.

Should you use Namespace?

A good fit if you're…

Teams on GitHub Actions who also build with Bazel and want remote cache and execution from the same vendor
iOS and macOS teams that need Apple-silicon runners and Mac dev environments
Teams that want runner shape, caching and privileges defined in the workflow's runs-on label rather than a dashboard

Probably not for you if you're…

Teams that shard tests very wide on a small plan: Team caps Linux at 64 concurrent vCPU
Windows CI on the free Developer plan, which has no Windows runners
Anyone who wants a single per-minute price per machine without unit-minute arithmetic

Pick something else if…

you only need faster GitHub Actions runners and want no concurrency cap and plain per-minute pricing → Blacksmith
your minutes fit the GitHub plan allowance and queueing is not a problem → GitHub Actions
builds must run on machines you own → Buildkite

Gotchas to know before you commit

Concurrency is a plan limit in vCPU, not jobs: 32 (Developer), 64 (Team) or 160 (Business) Linux vCPU, and 12, 24 or 48 on macOS. Jobs beyond it queue until capacity frees up (resource limits docs).
Unit minutes count RAM as well as vCPU: a shape costs max(vCPU, GB ÷ 2) units per minute, so a 4 vCPU, 16 GB runner bills twice a 4 vCPU, 8 GB one.
Queueing incidents are frequent. namespace-status.com lists 16 incidents titled elevated or increased queue time between August and 15 November 2026 (the latest on 11 November), plus control-plane error incidents on 29 October and in September.
Remote Docker builders are on by default for Namespace runners. Remove docker/setup-buildx-action from workflows or it overrides that configuration (GitHub Actions docs).
Plan limits differ: Windows needs Team or above, and jobs are capped at 3 hours on Developer and 5 on Team. GitHub's $0.002 per minute charge for self-hosted runners was postponed in December 2025 with no new date; if it returns, it would apply to third-party runners on Actions.

Namespace pricing

Checked Nov 15, 2026 against the official pricing page

Plans include unit minutes, where one unit is 1 vCPU and 2 GB RAM for a minute. Windows counts double and macOS ten times. Team is $100 a month for 100,000 units, and overage is $0.0015 per unit minute. It gets expensive with macOS and high-memory shapes, and when you need more concurrency, which means a higher plan.

Developer

$0 / month + usage

Trying it out, small Linux and macOS workloads

Pay per use for compute, Docker builds at $0.05 each
Linux and macOS runners, cross-run caching
30-day free trial, no credit card

Limit: Concurrency: 32 vCPU Linux, 12 vCPU macOS
Limit: No Windows runners; 3-hour job limit

Team

$100 / month

Small teams with steady CI

100,000 unit minutes
1,000 Docker builds
Windows runners

Limit: Concurrency: 64 vCPU Linux, 24 vCPU macOS
Limit: Overage $0.0015 per unit minute; $10 per 1,000 extra builds

Business

$250 / month

Larger teams, bigger machines

250,000 unit minutes
2,500 Docker builds
Larger shapes, RBAC, dedicated Slack channel

Limit: Concurrency: 160 vCPU Linux, 48 vCPU macOS
Limit: Same overage rates

Enterprise

Contact sales

Custom concurrency and volume discounts
Enterprise SSO, high-memory instances

Limit: Price not published

What it costs in practice

Team running 20,000 Linux minutes / month on 4 vCPU (8 GB)	$100 / month 4 vCPU × 8 GB = 4 units per minute; 20,000 × 4 = 80,000 unit minutes, inside Team's 100,000. Team's 64-vCPU cap allows 16 such jobs at once.
Same 20,000 minutes on 4 vCPU with 16 GB RAM	~$190 / month 16 GB counts as 8 units per minute (GB ÷ 2), so 160,000 unit minutes: $100 Team + 60,000 × $0.0015 = $190. Business ($250, 250,000 included) costs more unless usage grows. Team's 128 GB RAM cap also means only 8 of these jobs run at once.
iOS app: 2,000 macOS minutes / month on 6 vCPU (14 GB)	~$130 / month on Team 6 vCPU × 10 macOS multiplier = 60 units per minute; 2,000 × 60 = 120,000 units. $100 + 20,000 × $0.0015 = $130, if no other usage draws on the included minutes.

Estimates in USD from list prices. Your bill depends on usage, region, and discounts.

Health & activity

Updated Nov 15, 2026 · last 90 days

Closed sourceThe product itself isn't public. The figures below are for its public companion repo, so treat them as a partial view.

GitHub stars

181

24 forks

Commits

130

at least 10 committers

Merged pull requests

130

Issues opened / closed

1 / 0

Last push

Nov 14, 2026

Open issues & PRs

From namespacelabs/foundation (Home of the open-source nsc CLI (Apache-2.0)). Counts cover the last 90 days unless noted.

What the community says

Hacker News & GitHub issues · Aug 17, 2026 – Nov 15, 2026

Nearly all of the discussion is about one engineering post, "Mac at any scale" (3 November 2026), on how Namespace racks MacBook Pros in data centres to sell macOS capacity. Almost nobody discussed the GitHub Actions runners themselves in this window.

What people like

Engineering write-ups people read · 4 sources
The Mac post drew 421 points and 227 comments. Commenters called it an amazing article and asked for the promised follow-up on registry distribution.
HN thread 1 HN comment 2 HN comment 3

What people complain about

Laptop batteries in a data centre · 3 sources
Commenters questioned packing about a hundred MacBooks, and their lithium batteries, into each rack, and asked how the economics and cooling work compared with Mac minis.
HN comment 1 HN comment 2 HN comment 3
Licensing and supply questions · 4 sources
Some asked whether renting macOS VMs fits Apple's licence terms, and others objected to a provider buying up Mac hardware (and, per the post, an entire cable supply) to rent it out.
HN comment 1 HN comment 2 HN comment 3

Written by Tobias Brandt on Nov 15, 2026 after reading 16 relevant Hacker News comments, threads and GitHub issues from Aug 17, 2026 to Nov 15, 2026. Each point links to the discussions behind it, and the count shows how many sources raised it. This is a reading of public discussion, not a survey.