AOSplaneCompareCopilot
GitHub Copilot coding agent vs AOSplane
Assigning a GitHub issue to Copilot opens a pull request and asks you to review it. Copilot can do the same from a Linear issue. Opening the pull request is not the difference. AOSplane scopes, reviews, and QA-tests a Claude Code or Codex change before your team is the reviewer, and the pull request arrives with the spec, the review notes, QA’s results, and that task’s cost. Copilot draws on a plan’s AI credits. AOSplane stops a task that would cross its own cap. The steps and prices below are from GitHub’s docs as read on 11 November 2026.
- Hard spend cap per task
- Linear and Jira
- Claude Code and Codex
6 sections · 4 min read
On this page6 sections
The short version
If the team already lives in Copilot and is happy to be the first reviewer of its pull requests, assigning the issue is the short path. AOSplane is the path when that first review should already have happened, on Claude Code or Codex, whether the ticket was in Linear or Jira, and when the task has a hard spend cap instead of a balance of AI credits.
What assigning an issue does
GitHub’s kickoff doc is unambiguous: assigning an issue to Copilot always creates a pull request. Copilot receives the title, the description, and the comments that exist at that moment. Comments added on the issue after assignment are not passed along. Follow-ups go on the pull request. You can add an optional prompt, pick a base branch, pick a custom agent, and pick a model.
A prompt session is a different start. Copilot can work in a cloud sandbox, and you ask it to open a pull request, or you open one from the session when you are ready. Assigning an issue does not wait for that choice. The pull request is part of the assignment.
On AOSplane the draft waits for the loop you picked. With Plan first, you approve a spec before code is written. Review reads the change. QA runs the repository’s commands. The draft opens when those pass. Start now is the mode that does not wait for that loop. Either way, agents cannot push, and a person merges. The issue you started from can be a Linear issue or a Jira Cloud issue, not only a GitHub issue.
What the Copilot plan pays for
GitHub’s plans page says paid plans have access to Copilot in the cloud. Copilot Free does not. Organization owners and enterprise admins can turn features off, and the page says feature and model availability is subject to administrator policies and spending controls. Those controls exist. This page does not describe their screens.
Included usage is GitHub AI credits. The allowance table, read the same day, lists:
- Copilot Pro, $10 a month: 1,000 base credits and 500 flex, 1,500 in total.
- Copilot Pro+, $39 a month: 3,900 base and 3,100 flex, 7,000 in total.
- Copilot Max, $100 a month: 10,000 base and 10,000 flex, 20,000 in total.
- Copilot Business, $19 per granted seat a month: 1,900 credits per user, pooled.
- Copilot Enterprise, $39 per granted seat a month: 3,900 credits per user, pooled.
Usage past a Business or Enterprise pool is billed separately. Copilot access, for eligible enterprises, has no included credit allowance. Credits are GitHub’s meter. They are not an Anthropic invoice and not an OpenAI invoice.
AOSplane never resells the tokens. Claude Code calls your Anthropic account, or Bedrock, or Claude Platform on AWS, or another compatible provider. Codex calls your OpenAI account or Azure OpenAI. A Claude or ChatGPT subscription sign-in is rolling out and is always a personal account. Before the run, AOSplane reserves budget against the organization, the project, and the agent. Over the tightest remainder, the task does not start.
Side by side
| Copilot cloud agent | AOSplane | |
|---|---|---|
| Start | Assign a GitHub issue, or prompt a session on github.com. | Linear, Jira Cloud, the console, or a project API key. |
| Pull request | Assigning an issue always opens one. You are asked to review when it finishes. | Plan first opens a draft after review and QA pass. A person merges. |
| Follow-up | On the pull request. Later issue comments are not sent. | On a Linear issue, a reply to the AOSplane agent continues the session. Jira gets comments. |
| Bill | Copilot plan and AI credits. Admins have spending controls. | A hard cap checked before the task starts. The cost is part of what the team receives. |
| Proof | You review the pull request GitHub opened. | QA’s verdict is the command’s exit code. Review reads the diff. QA runs the checks. |
When Copilot is the better fit
- The work is already a GitHub issue, and the people who will review it live on that pull request.
- You want GitHub’s agent, GitHub’s model picker, and GitHub’s credit pool, without a second product.
- An admin already governs Copilot with the org’s policies and spending controls, and that is the control plane you want.
When AOSplane is the better fit
- The team should see a pull request that already has a spec, review notes, QA results, and a cost. Linear’s own agent and Copilot in Linear also open pull requests from issues.
- The harness should be Claude Code or Codex, with the same gate on either one.
- A task that would cross its cap should fail to start, rather than draw a shared credit pool down. Setting the Claude Code caps and the AOSplane cap.
- Several agents should share one loop. What changes when Claude Code is a team agent.
Assignment behavior is from GitHub’s kickoff doc. Plan prices and credit totals are from GitHub’s Copilot plans. Both were read on 11 November 2026 and may have changed. GitHub Copilot is not an AOSplane product.