Roadmap

This roadmap describes features planned or being considered for changAIs. Items may change as the product evolves.

Planned

AI Tool Integrations

The current release supports setup workflows for Claude Code, GitHub Copilot, Cursor, Windsurf, Aider, Gemini CLI, OpenCode, and Qwen Code.

Future integration work includes:

• Codex.
• MCP (Model Context Protocol) support, so compatible AI tools can access changAIs review context, report summaries, and reviewed-note state through a local integration.
• More AI coding assistants as their project-instruction and report-generation workflows stabilize.
• Direct report generation for more tools, so fewer workflows require copying a prompt manually.
• Richer feedback handoff between changAIs replies and AI tools.

Team Workflows

A team version of changAIs is planned for shared review across users.

Expected capabilities include:

• Reviewing pull requests with changAIs context.
• Sharing review state across teammates.
• Assigning or resolving notes as a team.
• Letting users add their own review notes, not only AI-generated notes.
• Tracking which changes and notes have already been reviewed.

Test and Lint Actions

changAIs should help verify files changed by AI and qualify AI notes against real project signals.

Planned actions include:

• Running relevant unit tests for changed files.
• Running integration or end-to-end tests connected to changed areas.
• Running lint checks for files changed by AI.
• Showing test and lint results near the review workflow.
• Qualifying notes and changed files with test and lint evidence, such as passed, failed, not covered, or not checked.
• Highlighting when an AI note claims behavior is correct but related tests or lint checks fail.
• Tracking a suspiciousness metric that increases when AI adds new packages, changes dependency manifests or lockfiles, or works in files and paths the user marked as unsafe or security-sensitive.

Time Tracking

Time tracking is planned inside the app so users can understand how much review time AI-generated changes require.

Potential views include:

• Time spent reviewing a report.
• Time spent per changed file.
• Time spent resolving notes.
• Review time trends across sessions.

AI Usage and Cost Tracking

changAIs may track how much AI work went into a change and its review, including token counts and estimated cost.

Potential views include:

• Tokens used to create the change.
• Tokens used to generate review notes.
• Tokens used while acting on replies or questions.
• Estimated cost per report, AI session, or commit.

Under Consideration

These ideas are not committed yet, but fit the direction of the product:

• Review checklist templates for security, tests, API behavior, migrations, configuration changes, and other recurring review concerns.
• Note categories such as risk, behavior change, test needed, style, cleanup, or unknown, so users can scan review work faster.
• Changed-file dependency context to show related callers, imports, tests, schemas, or configuration touched by an AI change.
• A regression watch list for sensitive areas such as auth, billing, permissions, migrations, deployment, and production configuration.
• Before and after behavior summaries for each report or changed file.
• Commit readiness state, such as needs review, needs tests, or ready to commit.
• Ignore rules for generated files, lockfiles, snapshots, build output, or project-specific paths.
• Session comparison across multiple AI runs.
• Local review history for reopening previous reports and decisions.
• Pull request description drafts based on changed files, reviewed notes, resolved issues, and test or lint results.
• Risk scoring for changed files based on size, touched areas, and note severity.
• Suspiciousness scoring based on dependency changes, new packages introduced by AI, sensitive file access, unsafe user-marked paths, and missing verification.
• Test coverage hints for files changed by AI.
• Exporting review summaries for pull request descriptions.
• Comparing reports across multiple AI sessions.
• Custom report paths and report history.
• Local-only settings for privacy-sensitive projects.
• Keyboard-first review navigation for moving quickly between files and notes.