Pimp My IDE / review pit wall
Back to garage
October 4, 2026 | code review / APIs / merge policy

An API call is not a review policy.

GitHub can now accept Copilot code review requests through REST and GraphQL. The useful move is not wiring every pull request to the new trigger. It is deciding which changes deserve which lane.

Automation may start the review. Change risk should set the effort. A human still owns the merge.

The trigger moved into code.

GitHub announced API support for requesting Copilot code review on October 2. A caller can use REST or GraphQL and set an effort level for each request. GitHub also changed its built-in default to Balanced on September 28. An explicit Lite setting stays Lite.[1]

This turns review into an operation that an issue tracker, release service, or internal tool can start. It also creates a policy question. A convenient endpoint can make review volume automatic before anyone decides which changes need deeper analysis.

Triggering review is transport. Routing review is judgment.

Effort has a job and a cost.

GitHub describes Lite as its standard review. Balanced applies deeper analysis to complex logic, security-sensitive code, and cross-service changes. Max appears in settings with a coming-soon label and is not available yet.[2]

The levels do not cost the same. GitHub currently estimates that one Lite review uses $0.05 to $1 in AI credits. Its estimate for Balanced is $0.25 to $5. Those ranges can change as models change, and they exclude GitHub Actions minutes.[3]

Do not turn those estimates into a price calculator. Use them to make one point visible. The deeper lane spends more. A repository default needs a budget owner, and an API caller needs a reason when it overrides that default.

The runner changes what the reviewer can see.

Copilot code review can use GitHub Actions for repository context and tool use. GitHub says a review still runs if Actions is unavailable, but without those agentic capabilities. If an organization disables GitHub-hosted runners, it can supply self-hosted runners or receive a more limited review.[3]

A review receipt should record that distinction. "Copilot reviewed this" hides whether the run had project context, which runner path it used, and whether a spending limit blocked the request. GitHub says reviews can be blocked when the relevant user, enterprise, or cost-center budget is exhausted.[3]

Keep the merge gate separate.

GitHub warns that Copilot may miss problems or make mistakes. Its documentation says to validate the feedback and add human review.[3] That is the line an automated trigger must not erase.

Route by the change in front of you. A small documentation edit may fit Lite. Authentication, billing, migrations, cryptography, infrastructure, and cross-service behavior belong in a deeper lane with a named human reviewer. Large generated changes may deserve reduction before review, not more automated commentary after the fact.

The pit lane below prepares a request card. It does not call GitHub, inspect a pull request, spend credits, run Actions, request a reviewer, or approve a merge.

Interactive makeover / review effort pit lane

Put each change in a lane.

This replaces one automatic review toggle with an effort selector, a change scope, four release requirements, and a copyable request card. It prepares policy. It does not submit a review.

Effort shifter

Choose one lane

The selected effort should match the failure cost and change shape. It does not measure review quality.

Review effort

Name the code boundary, not the ticket title.

Merge lane

Request draft
Request requirements
0 of 4 requirements selected

Prepare the scope before requesting an automated review.

Four selected requirements make the request structure complete. They do not prove that the API accepted it, Actions ran, the review found defects, a human checked the result, or the change is safe to merge.

Sources read

Source log and evidence boundary
  1. GitHub Changelog, "Copilot code review: API support and new default effort level", published October 2 and read October 4, 2026. This is the first-party source for REST and GraphQL request support, per-request effort, the Balanced default, the September 28 change, and the settings hierarchy.
  2. GitHub Docs, "Configuring code review by GitHub Copilot", read October 4, 2026. It defines Lite, Balanced, and the unavailable Max option. It also documents automatic-review sources, repository and organization settings, new-push review, draft review, and approval controls.
  3. GitHub Docs, "About GitHub Copilot code review", read October 4, 2026. It supports the runner behavior, limited-review fallback, estimated AI-credit ranges, separate Actions cost, budget blocking, attribution rules, and the requirement to validate feedback with human review.
  4. Hacker News discussion 49949235, read October 4, 2026 after resolving the exact item through the Hacker News API. It was part of the current-source scan and supplied discussion about hard caps versus service continuity. It does not verify GitHub product behavior.

Evidence boundary. GitHub documents its product behavior and current estimates. Pimp My IDE designed the risk-routing order and request card. This page does not use the GitHub APIs, connect to a repository, consume credits, run a review, or test review quality.