Self-hosted GitHub Actions
Dedicated runners for teams and solo developers using AI coding agents. Try one CI run free before you pay. Your agents connect the project. We maintain the server
Public website, customer portal with an admin area, and FastAPI backend
12:14
Six jobs waited for available memory. Five finished before the portal browser checks
A second project on the same server, a pull request check: 12:01 with 6,640 backend tests and 311 portal browser scenarios. Run time starts when the workflow launches and ends when the last job finishes. Your own measurement will show your result
After each change, your agent waits for the full CI result. A dedicated runner is worth testing if a run takes more than 15 minutes or jobs spend time in a queue.
Your agent can measure this right away: it reads the GitHub Actions run history and prepares a report. The report includes:
The agent only reads GitHub Actions data. It leaves code and settings unchanged
Get the CI measurement promptYour agent measures your current CI right away. Send us the report. We arrange one free run on our runner and compare the time with your CI. We agree access and data cleanup before the trial
After the trial, we agree the scope and fixed price. You choose a plan and pay the invoice. For the own-server plan, rent a server from a hosting provider and give us the agreed access
We prepare the server. Your agent follows the connection prompt to create a one-time token. You forward it to us so we can register the runners. The agent then moves heavy jobs to the server
Your agent repeats the measurement on the new server. It compares run time, test counts, retries and failures before and after connection
We rent the server, connect your CI and keep it running
$590
per month, USD
We set up CI for your first two repositories on your server
$990
one-time setup, USD
Optional support for the own-server plan
$330
per server per month, USD
Prices are in USD and include any taxes Progon must collect. We confirm the final itemized total, payment route and start date before you accept an order. Consumer cancellation rights apply as set out in the service terms
Additional repositories cost $65, $195 or from $390 to connect, depending on the work. We credit available included hours first
Connection is estimated at 3 hours. You have 2 included hours left this month, so we credit them first. You pay $65 once for the remaining hour. The monthly fee stays at $590
Heavy stages reserve memory before they start and release it when they finish
32 GB memory · SSD
No fixed CPU quotas per job
Memory is reserved before start
Projects run alongside each other
The next job can use the reservation
The same run shown in the timeline above
Memory, GiB
CPU threads in use
Drag the slider left or right
1. Screenshot checks started as soon as the backend build finished
| Time | Reserved, GiB | Used with cache, GiB | Processes, GiB | Threads |
|---|---|---|---|---|
| 0:01 | 6.4 | 18.3 | 0.8 | 0 |
| 1:02 | 21.2 | 19.7 | 2.3 | 2 |
| 2:02 | 25.8 | 22.8 | 3.8 | 10 |
| 3:02 | 25.6 | 23.2 | 3.8 | 9 |
| 4:03 | 25.6 | 25.4 | 7.0 | 24 |
| 5:04 | 25.6 | 25.0 | 6.8 | 22 |
| 6:04 | 26.0 | 24.8 | 7.3 | 26 |
| 7:00 | 26.0 | 24.5 | 7.4 | 26 |
| 8:01 | 26.0 | 25.1 | 7.8 | 25 |
| 9:02 | 26.0 | 23.8 | 6.8 | 23 |
| 10:02 | 24.7 | 19.6 | 2.8 | 10 |
| 11:02 | 22.2 | 17.6 | 1.6 | 4 |
| 12:03 | 6.4 | 15.8 | 0.9 | 0 |
When one project is idle, another can use the available memory and all CPU threads.
To add a project, your agent creates a token. We register its runners and configure the job stages. You do not need to write a separate resource allocator
Two projects ran their full checks on the same server at once. One finished in 13:36
Compare who runs the server and how costs change as your CI workload grows
| Option | Who manages it | Cost | As you run more CI |
|---|---|---|---|
| Progon managed server | We manage a dedicated server for your private Linux workloads | $590/monthServer, initial setup and support included | Same server feeWithin the agreed server capacity and work scope |
| GitHub-hosted runners | GitHub manages the runners. You maintain your workflows | Per minuteBeyond the usage included in your GitHub plan | More billed minutesCharges depend on runner type and included usage |
| Your own server | You rent the server. We set up CI, with optional ongoing support | $990 onceHosting paid separately Optional support: $330/month | Capacity is the limitExtra resources or another server need a separate agreement |
GitHub Actions stays your workflow platform with every option. Progon runs heavy jobs on self-hosted runners. You can leave light jobs on GitHub-hosted runners
A fixed server fee covers the agreed capacity. More jobs may wait when memory is full. We assess your workload before connection and agree any extra resources in advance
We have not benchmarked the example suite against a standard GitHub-hosted runner. Use the free trial to compare with your current CI
Give the prompt to your agent. It needs repository access and the GitHub CLI. Use Claude Code, Codex, Cursor or another coding agent
Your agent prepares a report on your current CI right away. It only reads GitHub Actions data and leaves code and settings unchanged
Measure GitHub Actions CI for this repository right now, using the run history from the last 7 days. Do not change code or settings.
1. List runs from the last 7 days:
gh run list --limit 500 --created ">=DATE_7_DAYS_AGO" --json databaseId,workflowName,event,createdAt,conclusion
2. Get the jobs for each run:
gh api repos/{owner}/{repo}/actions/runs/RUN_ID/jobs --paginate
3. For each job, calculate queue time (started_at minus created_at) and execution time (completed_at minus started_at). Identify checkout, dependency installation, build and test steps separately.
4. Report the run count, total job minutes, median and 90th percentile of full run time, the share of time spent in the queue, and the five slowest steps.
5. Assess whether a Progon dedicated server fits this repository. It fits private repositories with Linux jobs when a full run takes more than 15 minutes or jobs wait in a queue. It does not support public repositories or macOS, iOS or Windows builds.
6. Save the report to ci-measurement.md, show it to me, and suggest that I send it to Telegram @progonci.
After payment, we prepare the server and send you the runner label
Move CI for this repository to a Progon dedicated server. Runner label: LABEL_FROM_PROVIDER.
1. Create a one-time runner registration token, valid for 1 hour:
gh api -X POST repos/{owner}/{repo}/actions/runners/registration-token --jq .token
For organization runners: orgs/{org}/actions/runners/registration-token
2. Give the token to me so I can forward it to the provider. Do not save it anywhere.
3. Wait for runners with this label to appear with online status:
gh api repos/{owner}/{repo}/actions/runners
4. In .github/workflows, change runs-on for heavy jobs (tests, builds and browser checks) to [self-hosted, LABEL_FROM_PROVIDER]. Light jobs can stay on ubuntu-latest.
5. Open a PR, wait for the run to finish, and compare its duration with ci-measurement.md.
6. Do not change repository secrets or permissions.
Message us on Telegram to arrange one free run on our runner