Risk-adaptive multi-agent PR review for Claude Code & Cursor — 7 specialized agents catch bugs, security flaws, and AI hallucinations in parallel
# Soliton
[](LICENSE)
[](https://github.com/andyzengmath/soliton)
[](https://cursor.com/marketplace)
[](agents/)
**Your AI coding agent writes code 10x faster than you can review it. Soliton reviews it for you.**
A Claude Code & Cursor plugin that runs 2-7 specialized review agents **in parallel**, adapting to how risky the PR actually is. Low-risk README fix? Two quick agents, done in seconds. Critical auth refactor? Full 7-agent battery with security data-flow analysis and hallucination detection.
> **Assist Mode** — every review surfaces high-confidence findings organized by severity plus a risk score, but makes no automated merge/reject decisions. You stay in control.
## Why Soliton?
| Problem | How Soliton Helps |
|---------|-------------------|
| AI agents generate code faster than you can review it | Parallel multi-agent review catches issues in seconds, not hours |
| Agent-generated code hallucinates APIs and wrong signatures | Dedicated hallucination detector verifies every new import and function call |
| Generic review tools waste time on low-risk changes | Risk-adaptive dispatch: simple PRs get 2 agents, complex ones get 7 |
| Review findings are vague ("consider improving...") | Every finding includes confidence score, exact location, and concrete fix code |
| No way to feed review results back to coding agents | `--feedback` mode outputs machine-consumable instructions your agent can execute |
## Quick Start
```bash
# Step 1: Add the marketplace (once)
/plugin marketplace add andyzengmath/soliton
# Step 2: Install the plugin
/plugin install soliton@soliton
# Review your current branch
/pr-review
# Review a GitHub PR
/pr-review 123
```
That's it. Soliton aAgent that generates comprehensive documentation, API references, architecture diagrams, and developer onboarding guides from existing code.
Agent configuration for systematic bug investigation that traces issues from error logs through the codebase to root cause with suggested fixes.
Agent for integrating third-party APIs including SDK setup, type generation, error handling, retry logic, and rate limit management.
Cursor's built-in autonomous coding agent that can make multi-file edits, run terminal commands, search the codebase, and iteratively build features with minimal human intervention.
Cloud-based autonomous coding agent that runs in the background on remote sandboxed environments, handling complex multi-step tasks while you continue working.
Cursor's multi-file editing agent within Composer mode that can create, edit, and delete files across your entire project in a single conversation.