Enterprises working with a GitHub or Microsoft account team can get Copilot included for every GitHub Enterprise licensed user. Usage is metered against an enterprise budget with no per-seat monthly credits. Self-serve customers get it later, in stages.
How we checked
GitHub's own changelog read; availability, $10 Standalone price and budget behavior confirmed
Why it matters
Copilot cost shifts from seat counts to usage budgets. Completions keep working when the budget runs out, but chat and cloud agents stop.
Anthropic expanded its AI for Science program, which gives free Claude credits, to 20 foundations, institutions and research projects. These include the Sloan Foundation, the US NSF, EMBL, STScI, the Broad Institute's gnomAD and the Tau Ceti Lean math library.
How we checked · Why it matters
How we checked
Fetched Anthropic's announcement page; partner list and 20-partner count confirmed
Why it matters
Researchers may be able to get Claude credits through funders and institutions they already work with, not only by applying directly.
GitHub Agentic Workflows is generally available for Free, Team and Enterprise. You write workflows in Markdown and they compile to standard Actions YAML. Agents run read-only by default in a sandboxed container behind a firewall, and proposed changes get a threat scan.
How we checked · Why it matters · Try this
How we checked
Changelog read; github/gh-aw repo (MIT, install command) and docs site confirm Markdown-to-Actions compile; not run
Why it matters
You can run agent jobs like issue triage or CI failure analysis inside your existing Actions setup and policies, with safeguards turned on by default.
Try this
In a throwaway repo, run gh extension install github/gh-aw, add one triage workflow, and read the compiled Actions YAML before enabling it.
Shared Copilot Automations are rolling out. They belong to the repository, are editable by anyone with write access, act as Copilot, and bill AI credits to the repo owner. Org repos need an admin policy enabled first.
How we checked · Why it matters · Try this
How we checked
GitHub changelog entry read; rollout status, billing and permissions confirmed there
Why it matters
Team agent workflows like bug reproduction keep running after the creator leaves. Credits go to the org, so admins need to watch spend.
Try this
Ask an org admin to review the "Allow use of Copilot cloud agent billed to the organization" policy before creating a shared automation.
GitHub Copilot Fusion is generally available. It routes each prompt across models using Single, Cascade or Critique workflows, with Standard, High and Max effort levels. Each model used is billed at its own rate, with a temporary 30% discount.
How we checked · Why it matters · Try this
How we checked
GitHub changelog read; GitHub docs page mentions Copilot Fusion in the model picker; no hands-on test
Why it matters
You can stop picking a model per task. Admins can limit which models Fusion uses, and costs depend on whichever models it picks.
Try this
In VS Code or Copilot CLI, choose Copilot Fusion Standard in the model picker for one day of routine tasks, then compare AI credit use with your usual model.
A GitHub security extension for the Copilot app, in public preview, lists open code scanning alerts by severity and lets Copilot explain or fix them. Fixes go to your local working copy only. A /security-review command checks uncommitted changes.
How we checked · Why it matters · Try this
How we checked
GitHub changelog entry read, including install command and plan requirements; not run
Why it matters
You can triage and fix security alerts in one place, and nothing is committed or pushed until you review it.
Try this
Install it with copilot plugin install github-security@awesome-copilot, then run /security-review on a small local change.
Copilot Memory is generally available on paid Copilot plans and keeps preferences and repo conventions across sessions. Copilot dreaming, an opt-in public preview, updates memories and proposes changes to skills and instruction files through pull requests.
How we checked · Why it matters · Try this
How we checked
GitHub changelog entry read; GA and preview status and opt-in rules confirmed there; not tested
Why it matters
Agents stop relearning your conventions every session. Dreaming uses AI credits and edits memories automatically, so review what it stores.
Try this
Open your Copilot settings and review or delete saved personal memories before you turn on dreaming.
OpenAI says eligible ChatGPT Pro subscribers can claim $100, $200 or $500 per month in API credits by tier. Business plans get $20 per Standard seat or $100 per Premium seat, pooled up to $500 per workspace.
How we checked · Why it matters
How we checked
Only OpenAI's own video description read; help.openai.com article and openai.com returned 403, so the terms were not confirmed
Why it matters
If this holds, existing ChatGPT subscribers can prototype API apps and agents without separate API spend.
At GitHub Universe, GitHub's Jay Parikh highlighted Octoverse 2026 figures as reported by Simon Willison: merged PRs tripled to 220 million per month, commits reached 7.38 billion per month, and new repos reached 30.3 million per month.
How we checked · Why it matters
How we checked
GitHub's Octoverse article gives 12-month totals (1.3B merged PRs, 34.3B commits), not the monthly figures; those were not confirmed
Why it matters
These are headline numbers for how much agents are adding to GitHub activity, but the published report frames them differently.