CloudAEye Code Review — MCP Server¶
An MCP server that gives your coding agent (Claude Code, Cursor, OpenAI Codex, or any MCP client) a set of pre-commit code-review tools. It inspects code you have not yet committed for issues such as bugs, task misalignment, security problems, and more — all before the code leaves your machine.
Documentation¶
| Page | What it covers |
|---|---|
| Overview | What the server is, the review model, and the command surface |
| Setup | Connect the MCP server, install the skills, configure environment |
| Skills & Commands | The slash commands your agent runs: Inspect, Describe, Check Task, and more |
| Custom Context | Code-graph refresh, tenant keys, and Jira / GitHub ticket resolution |
| Tool Reference | HTTP endpoints and MCP tool signatures |
Command surface at a glance¶
Every command follows the same / slash-command convention in each IDE.
| MCP tool | Claude | Codex | Cursor | Purpose |
|---|---|---|---|---|
inspect_diff |
/inspect |
/inspect |
/inspect |
Single-shot bug + security-essentials pass over the pending diff |
describe_change |
/describe |
/describe |
/describe |
Plain-markdown summary of what the change does |
check_task |
/check_task |
/check_task |
/check_task |
Verify the diff against a task, ticket, or spec (DONE / PARTIAL / NOT DONE) |
security_review |
/security_review |
/security_review |
/security_review |
Security-focused review across all security report types |
review |
/review |
/review |
/review |
The full review — every report type and category |
Core ideas¶
- Pre-commit. The unit of review is the complete change you are about to commit — including brand-new files you haven't added to the repository yet, not just edits to tracked files — compared against your project's main branch selected during GitHub integration. Everything is reviewed before it lands.
- Session-based. A review session is keyed by
(repo, branch, head). The diff is uploaded to the server and lives there, keyed by session id — it is never passed inline to a tool. - Agent-native. Findings come back as structured data your coding agent acts on — it keeps calling the tools until the issues are resolved.