Actions / Local & hosted

Your compute. Or ours. One clear choice.

Keep execution on your organization’s devices with a fixed contributor seat. Choose separately approved hosted capacity when you need it, billed by usage and protected by spending limits.

$15 USDper contributor / month

Your 15-day trial starts after organization approval. No automatic charge. Hosted usage is separate from the local trial.

your organizationWalkthrough
Client workspace01 / 04

Finish organization setup

The overview shows the next incomplete setup step.

Tbench / organization
Overview
✓ Organization approved
✓ GitHub installation connected
Next · confirm contributor access
Open execution when setup is ready
Step explained
Example workspace

01 / Local runners

$15 USD

per contributor / month

Annual billing

$180 USD per contributor / year

Billed once for 12 months. Same seat rate.

Taxes and final payment currency are confirmed before payment. Your bank or payment provider may apply currency conversion and fees.

Try free for 15 days

Your 15-day trial starts after organization approval. No automatic charge. Your approved quote confirms monthly or annual terms before payment.

Every seat includes.

  • Explicit contributor seats across both execution paths
  • Organization-bound repository and device access
  • GitHub workflows and the independent Tbench runner protocol
  • Separate local application-verification command
  • Team roles, enrollment and revocation controls
  • Execution activity and a separate billing workspace
  • No per-minute Tbench charge for your own compute

02 / AI runner demo

One run. One clear price.

$0.10USD/ run

Approved Linux x64 CPU compute included: up to 2 CPU / 4 GiB and ten minutes including preparation and cleanup. No additional compute-minute charge.

Introductory demo pricing requires operator approval, an approved workload and qualified capacity. Rates exclude tax. Paid model APIs, managed cache and hosted benchmark audits are separate.

Every run has a ceiling.

  • Monthly organization budget and per-run cost limits
  • Fixed fee reserved before enqueueing
  • No charge when cancelled before the run is leased
  • One fixed charge for a started run, including failures
  • Uncertain jobs retain their reservation for review
  • Local jobs never consume hosted credits

01 / Seat definition

Count people once.

Each approved, active owner, admin or developer you explicitly enroll occupies one organization seat, whether they use GitHub workflows or the independent runner. Viewers and billing-only accounts are excluded. Add approved teammates by email in your workspace; repository collaborators are never silently imported or billed.

02 / Your infrastructure

Bring the capacity.

Electricity, hardware, Docker licensing, GitHub charges and optional model inference are separate. Local execution can reduce hosted minutes, but savings depend on utilization and maintenance.

03 / Payment

A clear quote for your team.

Your organization owner or billing manager requests a quote in the workspace. Confirm seats, currency, taxes and your monthly or annual term before payment.

Monthly workload scenario

10M jobs.
A different cost equation.

Model a team’s CI spend with published runner rates and a flat contributor fee. Keep execution on the capacity your organization owns.

1MPull requests in this scenario
10MJobs at 10 per pull request
$298,500Monthly fee gap vs GitHub, before local compute costs

GitHub-hosted

$300,00050M minutes × $0.006

Blacksmith

$200,00050M minutes × $0.004

Tbench Actions

$1,500100 contributors × $15 · plus your compute costs

Scenario inputs: 1M pull requests × 10 jobs × 5 billable minutes, with 100 contributors for one month. Linux x64, 2-vCPU list rates; provider memory allocations vary. Fees shown before free allowances, negotiated discounts, taxes, hardware, electricity, storage and operations. Execution capacity is scoped during onboarding.

The same scenario has a $198,500 monthly fee gap against Blacksmith before local compute costs. Use your own job durations, contributor count and operating costs when assessing your workload.

Your plan at a glance

Know what your seat covers.

Organization access and execution on your enrolled devices, with optional services scoped separately.

Actions seat scope
ServiceStatus
Your organization’s enrolled devicesCompatible workloads, subject to approved scope and device capacity
First 15 daysApproved organization trial for Actions and local application verification. Your organization supplies compute; paid terms require a separate decision.
Hosted computeSeparate pay-as-you-go plan, operator-approved capacity and explicit organization budgets. Native GitHub jobs remain on the existing enrolled-device path.
Persistent managed cacheNot included in either runner plan
LLM reviewOptional repository-defined workload; local inference and its egress policy must be configured separately
Benchmark audit and certificationSeparate methodology, acceptance criteria and any approved hosted-run allowance; contributor seats are not model credits
Automatic subscription or renewalNo mandate is created by an access or payment request

Your approved quote confirms the seat count, final currency, taxes and payment terms. The billing workspace tracks payment requests and provider-confirmed status.

Before you get started

A few practical details.

How seats, execution and payment work for your organization.

When does the free trial start?

The 15-day trial starts when Tbench approves your organization, not when you submit the form. The same repository, device and policy approval checks apply. There is no automatic charge or trial reset on reapproval.

How are contributor seats counted?

Each explicitly enrolled, approved and active owner, admin or developer counts once across execution paths. Viewers and billing-only accounts are excluded. GitHub access is verified separately. Your billing quote confirms the seat count.

Can a local command run work on another teammate’s device?

The contributor-local push command targets the developer’s own opted-in computer. Hosted execution uses your organization’s separate capacity and budget settings.

Where does workflow data go?

Execution stays on your selected device. GitHub handles workflow orchestration, logs and configured artifacts on its integration path. Tbench’s independent queue handles approved policy and result metadata; exact source commits are prepared through your organization’s own Git access. Optional model review has its own inference and network configuration.

How does payment work?

Start by requesting a quote in your billing workspace. Stripe checkout follows merchant activation and approved billing details. Until then, the workspace accepts billing review requests. A free trial does not authorize renewal.

15 days to put it to work

Bring your organization’s next workflow.

Apply with your company email. We review your organization, repositories and device scope. Your 15-day trial starts after organization approval. No automatic charge.

Try free for 15 days