GitHub Copilot Moves to Usage-Based Billing...

Correction — June 1, 2026:

When this article was first published May 27, 2026, I asserted that Microsoft had not published Copilot's per-Credit USD cost, per-feature consumption rates, or hard cap mechanism. All three were already published. The per-Credit rate (1 AI Credit = $0.01 USD) was in GitHub Docs since at least April 27, 2026, the day GitHub announced the transition. Per-token model breakdowns were in the same documentation. The hard cap mechanism — a user-settable $0 overflow budget that makes additional charges impossible — was documented and publicly discussed in GitHub's billing reference.

I missed all three. The original framing ("Microsoft has not published" / "the meter rate is unknown") was empirically wrong. The corrected analysis below uses the actual published rates and identifies the substantive consumer-protection concern that was always the real story: GitHub published transparent rates, but those rates predict that typical agentic Copilot workflows will exhaust monthly Credit allowances in a single working session, producing structural overage for any meaningful agentic user.

Original text is retained in this article's git history. The retraction is logged on the OneHuman Record page as the founding correction entry.

The number

$0.01 per AI Credit. That is the published per-Credit rate for GitHub Copilot's usage-based billing structure, effective today, June 1, 2026. Pro subscribers receive 1,000 Credits per month at the $10 plan price. Pro+ receives 3,900 Credits at $39. Business and Enterprise receive $19 and $39 of Credits per seat respectively. The rates are public, in GitHub Docs, since April 27. The math is the consumer-protection story.

What changed, plain English

Prior structure (through May 31, 2026): Flat subscription price. Premium models consumed a "premium request" allowance included in your plan. Standard chat was effectively free within the subscription.

New structure (June 1 forward): Same subscription price as the base. Every feature — autocomplete, chat, agent mode, premium models — draws from a monthly GitHub AI Credits allowance at $0.01 per Credit. Usage beyond the allowance is billed at the same per-Credit rate.

The conversion: plan price in USD = monthly Credits. $10 Pro = 1,000 Credits. $39 Pro+ = 3,900 Credits. $19 Business = 1,900 Credits per seat. $39 Enterprise = 3,900 Credits per seat.

The math Microsoft disclosed but under-communicated

A typical agentic Copilot session — using Workspace, Agent Mode, or large-codebase refactoring with premium models — consumes between $30 and $40 in token costs at the published per-token rates. That is 3,000 to 4,000 Credits per session.

The math:

Pro subscriber ($10/month, 1,000 Credits): a single agentic session exhausts 25-33% of the monthly allowance. Two sessions exhaust the allowance entirely. Three sessions means overage or cutoff.
Pro+ subscriber ($39/month, 3,900 Credits): approximately one full agentic session per month before overage.
Business team lead ($19/seat, 1,900 Credits per seat): a 20-developer team with moderate agent usage hits structural overage within the first week of the month.

The base prices ($10, $39, $19, $39) are now floors, not ceilings. The published rates predict that every meaningful agentic Copilot user is structurally over allowance every month. The disclosure was honest. The implication was under-communicated. The watchdog story is the math, not the secrecy.

The $0 overflow budget option you probably don't know about

GitHub provides a user-settable overflow budget option, including $0 (no overage allowed). Setting this to $0 makes additional charges impossible — Copilot stops working for the month once the allowance is exhausted. This prevents bill shock at the cost of interrupted workflow.

Most subscribers will not know this option exists unless they specifically read the billing documentation. Set it before the first agentic session.

Consumer protection Q&A

Q: Is there a hard cap on overage charges?

A: Not a hard cap imposed by GitHub, but a user-controlled one. Set the overflow budget to $0 in your billing settings. That creates the equivalent of a hard cap — usage stops when the allowance runs out. Unset, the meter runs.

Q: What does Copilot cost if I use Agent Mode regularly?

A: At $30-40 per typical agentic session and a Pro allowance of 1,000 Credits ($10 equivalent): meaningful agentic use generates overage from the first or second session each month. Pro+ ($39/month, 3,900 Credits) covers one full agentic session. If you use Agent Mode more than once a month with any frequency, model your actual session costs against your allowance before assuming your bill is unchanged.

Q: How does Copilot compare to alternatives under the new structure?

A: Cursor at $20/month provides a usage allowance with predictable per-tier pricing. Claude Code at the Max $100/month tier provides 5-20x rate limits at a flat price with no metered billing. For heavy agentic use where predictable monthly cost matters, both are more transparent under the new Copilot structure. For light use (autocomplete, occasional chat), Copilot's $10 entry price remains the lowest in the category.

Q: Were existing subscribers grandfathered?

A: No. All paid subscribers migrated to AI Credits on June 1, 2026, regardless of original subscription date.

What to do today

Before the first agentic session: Log into GitHub billing settings, locate the overflow budget option, set it to $0. This prevents any bill above your base subscription price.

Model your usage: Identify whether your workflow is primarily autocomplete (likely within allowance) or Agent Mode/Workspace (likely structural overage). If you use agentic features more than once a month, your June bill will reflect it.

Evaluate alternatives if needed: Cursor, Claude Code, or Codex free tier are the comparison set for agentic coding cost predictability. The compare page at OneHuman reflects the current pricing across all four.

The OneHuman verdict

Conditional recommendation requiring user-by-user cost modeling. Light users (autocomplete only) stay within allowance and see no bill change. Agent Mode and Workspace users at any meaningful frequency will exhaust monthly allowance and should set the $0 overflow budget option before the first session to prevent surprise bills, then accept that Copilot pauses when allowance exhausts.

Heavy agentic users should evaluate Cursor ($20/month flat) or Claude Code ($100/month Max tier) as alternatives where the cost structure for sustained agentic workflows is more predictable. The base price comparison ($10 Copilot vs $20 Cursor) no longer reflects the real monthly cost for agentic use.

Sources and dates

GitHub Docs — Copilot Billing Models and Pricing (docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing, accessed June 1, 2026)
GitHub Blog — "GitHub Copilot is moving to usage-based billing" (April 27, 2026)
Find Skill.ai — "GitHub Copilot AI Credits Billing June 2026" analysis (May 31, 2026)
Enterprisedna — "GitHub Copilot Usage-Based Billing Enterprise 2026" analysis
OneHuman pricing sweep persistence layer entries 2026-05-24 (corrected June 1, 2026)
Cross-reference: AI Tool Pricing Comparison
Cross-reference: BNA #1 — OpenAI ChatGPT Ads Manager
Cross-reference: BNA #3 — Grok $300/Month Under SpaceX

Verified by OneHuman · June 1, 2026 (corrected from original May 27, 2026 publication)

Independent. AI-assisted. Human-verified. No ads. No affiliates. No investors.

GitHub Copilot Moves to Usage-Based Billing June 1, 2026 — The Math Predicts Structural Overage for Agentic Users