https://github.com/axect/magi-researchers

Three AI models, one synthesis — Claude, Gemini & Codex cross-verify each other for rigorous multi-perspective research
https://github.com/axect/magi-researchers

ai-research claude-code claude-code-plugin codex cross-verification gemini mcp multi-model research-automation scientific-computing

Last synced: 2 days ago
JSON representation

Three AI models, one synthesis — Claude, Gemini & Codex cross-verify each other for rigorous multi-perspective research

• Host: GitHub
• URL: https://github.com/axect/magi-researchers
• Owner: Axect
• License: mit
• Created: 2026-02-18T11:05:28.000Z (10 days ago)
• Default Branch: main
• Last Pushed: 2026-02-23T07:46:01.000Z (5 days ago)
• Last Synced: 2026-02-23T15:56:36.622Z (5 days ago)
• Topics: ai-research, claude-code, claude-code-plugin, codex, cross-verification, gemini, mcp, multi-model, research-automation, scientific-computing
• Size: 257 KB
• Stars: 9
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • Contributing: CONTRIBUTING.md
  • License: LICENSE

Awesome Lists containing this project

README

          


  



MAGI Researchers




  Three AI models. One synthesis. Zero lost progress.


  Multi-model research pipeline for Claude Code — Claude, Gemini, and Codex debate, cross-verify, and synthesize publication-ready artifacts.





   

   

   

   

  





  Get Started •

  Why MAGI? •

  Features •

  Usage •

  Roadmap •

  Changelog



---

> *Like the MAGI system in Evangelion — three supercomputers cross-verifying each other — this plugin orchestrates Claude, Gemini, and Codex for rigorous, multi-perspective research.*

## Get Started

**Prerequisites:** [Claude Code](https://docs.anthropic.com/en/docs/claude-code) + Python 3.11+ with [uv](https://docs.astral.sh/uv/) + [Gemini CLI](https://github.com/google-gemini/gemini-cli) + [Codex CLI](https://github.com/openai/codex)

**1. Install the plugin** (inside Claude Code):

```

/plugin marketplace add Axect/magi-researchers

/plugin install magi-researchers@magi-researchers-marketplace

```

**2. Set up MCP servers** (one-time):

```bash

claude mcp add -s user gemini-cli -- npx -y gemini-mcp-tool

claude mcp add -s user codex-cli -- npx -y @cexll/codex-mcp-server

claude mcp add -s user context7 -- npx -y @upstash/context7-mcp@latest

```

**3. Run your first research:**

```

/magi-researchers:research "your research topic" --domain physics

```

MAGI generates cross-verified hypotheses, writes implementation code, renders publication-quality plots, and synthesizes a structured report — all saved to `outputs/{topic}/`.

Alternative: Local Development

```bash

git clone https://github.com/Axect/magi-researchers.git

claude --plugin-dir /path/to/magi-researchers

uv add matplotlib SciencePlots numpy

```

## Why MAGI?

Single-model research has blind spots. One model hallucinates a citation or misses a critical constraint — and nobody catches it.

| | **Single Model** | **MAGI (3 Models)** |

|:---|:---|:---|

| **Brainstorming** | One perspective | Three independent perspectives |

| **Verification** | Self-review (unreliable) | Cross-model peer review |

| **Blind spots** | Undetected | Caught by competing models |

| **Output** | Raw text | Structured report with consensus & divergence analysis |

- **Claude** — *The Scientist.* Synthesis, planning, implementation, report generation.

- **Gemini** — *The Critic.* Creative brainstorming, cross-verification, broad knowledge.

- **Codex** — *The Builder.* Feasibility analysis, code review, implementation focus.

## Features

### Research Pipeline

| Phase | What Happens | Output |

|:---|:---|:---|

| **1. Brainstorm** | Three models generate and cross-review ideas with expert personas | `brainstorm/` |

| **2. Plan** | Concrete research plan, stress-tested by a hostile reviewer | `plan/` |

| **3. Implement** | Claude writes code with live library doc lookups | `src/` |

| **4. Test & Visualize** | Collaborative test design + publication-quality plots | `tests/` + `plots/` |

| **5. Report** | Structured report with cross-verified claim-evidence integrity | `report.md` |

### Never Lose Your Work (v0.4.0)

Long research sessions crash. Context windows expire. Networks drop. Now you can pick up right where you left off:

- **`--resume`** — Interrupted mid-pipeline? Just pass `--resume ` and MAGI detects your progress from existing files. No manual bookkeeping, no fragile state files — your artifacts *are* the checkpoints.

- **Artifact contracts** — Before each phase, MAGI verifies that all required upstream files actually exist and aren't empty. Catches silent failures before they cascade into garbage outputs.

- **Standalone phase gates** — Every sub-skill (`/research-implement`, `/research-test`) now generates its own quality gate, even when run independently outside the main pipeline.

### Stress-Tested by Design (v0.3.0)

- **Weighted direction scoring** — Rank research ideas by novelty, feasibility, impact, rigor, and scalability with domain-tuned or custom weights

- **Dynamic persona casting** — Each model gets a topic-specific expert identity (e.g., *"Bayesian statistician with causal inference expertise"*), sharpening ideation

- **Adversarial debate** — At `--depth high`, models defend, concede, or revise on their top disagreements before synthesis

- **Murder board** — Gemini attacks the research plan as a hostile reviewer; Claude documents mitigations for every flaw found

- **Phase gates** — Automated quality checkpoints with conditional MAGI mini-review before each user approval step

- **Depth control** — `--depth low` for fast/cheap runs, `medium` (default) for standard review, `high` for full adversarial analysis

### Core Capabilities

- **Publication-quality plots** — `matplotlib` + `scienceplots` (Nature theme), PNG 300 dpi + vector PDF

- **MAGI traceability review** — All three models cross-verify the final report for orphaned claims and figures

- **Report gap detection** — Auto-generates missing visualizations from existing data

- **Domain templates** — Built-in context for Physics, AI/ML, Statistics, Mathematics, and Paper Writing

- **Journal strategy** — Venue recommendations for [Physics](docs/journal-strategies.md#particle-physics-phenomenology), [AI/ML](docs/journal-strategies.md#aiml-conferences--journals), and [Interdisciplinary](docs/journal-strategies.md#interdisciplinary-science-ml--natural-sciences) research

- **LaTeX math** — Proper inline and display equations across all outputs

Under the hood

- **Plot manifest** — Structured `plot_manifest.json` with metadata, section hints, and captions for automated report integration

- **Gemini fallback chain** — Resilient 3-tier model fallback: `gemini-3.1-pro-preview` → `gemini-3-pro-preview` → `gemini-2.5-pro`

- **Cross-phase artifact contracts** — Each phase validates incoming artifacts before running (tool-based Glob/Read, not LLM guesswork)

- **Depth-controlled token budget** — `--depth low` skips cross-review for fast/cheap runs; `--depth high` enables full adversarial debate

## Usage

| Command | Description |

|:---|:---|

| `/magi-researchers:research "topic"` | Full pipeline (all 5 phases) |

| `/magi-researchers:research-brainstorm "topic"` | Brainstorming with cross-verification |

| `/magi-researchers:research-implement` | Implementation (needs existing plan) |

| `/magi-researchers:research-test` | Testing & visualization |

| `/magi-researchers:research-report` | Report generation |

### Flags

| Flag | Values | Default | Description |

|:---|:---|:---|:---|

| `--domain` | `physics` `ai_ml` `statistics` `mathematics` `paper` | auto-inferred | Research domain for context and weight defaults |

| `--weights` | JSON object | domain default | Custom scoring weights (keys: `novelty`, `feasibility`, `impact`, `rigor`, `scalability`) |

| `--depth` | `low` `medium` `high` | `medium` | Review thoroughness — controls cross-review and adversarial debate |

| `--resume` | `` | — | Resume an interrupted pipeline from the last completed phase |

```bash

# Quick brainstorm with default settings

/magi-researchers:research "neural ODE solvers for stiff systems" --domain physics

# Deep analysis with custom weights and adversarial debate

/magi-researchers:research "causal inference in observational studies" --domain statistics --depth high --weights '{"novelty":0.2,"rigor":0.4,"feasibility":0.2,"impact":0.15,"scalability":0.05}'

# Resume a crashed session — MAGI picks up where you left off

/magi-researchers:research "neural ODE solvers" --resume outputs/neural_ode_solvers_20260225_v1

# Fast ideation only (no cross-review, lowest cost)

/magi-researchers:research-brainstorm "transformer alternatives for long sequences" --domain ai_ml --depth low

```

> If MAGI saves you research time, consider leaving a [star](https://github.com/Axect/magi-researchers/stargazers) so other researchers can find it.

### Output Structure

```

outputs/{topic_YYYYMMDD_vN}/

├── brainstorm/       # Personas, ideas, cross-reviews, debate, synthesis

├── plan/             # Research plan, murder board, mitigations, phase gate

├── src/              # Implementation + phase gate

├── tests/            # Test suite + phase gate

├── plots/            # PNG + PDF + plot_manifest.json

└── report.md         # Final structured report

```

Each phase produces artifacts that double as resume checkpoints — just pass `--resume` to continue from where you left off.

Full artifact tree

```

brainstorm/

├── weights.json              # Scoring weights

├── personas.md               # Expert personas

├── gemini_ideas.md           # Gemini brainstorm

├── codex_ideas.md            # Codex brainstorm

├── gemini_review_of_codex.md # Cross-review (depth ≥ medium)

├── codex_review_of_gemini.md # Cross-review (depth ≥ medium)

├── debate_round2_gemini.md   # Adversarial debate (depth = high)

├── debate_round2_codex.md    # Adversarial debate (depth = high)

└── synthesis.md              # Weighted synthesis

plan/

├── research_plan.md          # Research plan

├── murder_board.md           # Plan stress-test

├── mitigations.md            # Flaw mitigations

└── phase_gate.md             # Plan quality gate

src/

├── *.py                      # Research implementation

└── phase_gate.md             # Implementation quality gate

tests/

├── test_*.py                 # Test suite

└── phase_gate.md             # Test quality gate

```

Recommended Permissions

Add to `.claude/settings.local.json`:

```json

{

  "permissions": {

    "allow": [

      "Bash(uv:*)",

      "Bash(uv run:*)",

      "Bash(uv run python3:*)",

      "Bash(uv add:*)",

      "Bash(uv sync:*)",

      "Bash(mkdir:*)",

      "mcp__gemini-cli__ask-gemini",

      "mcp__gemini-cli__brainstorm",

      "mcp__codex-cli__ask-codex",

      "mcp__codex-cli__brainstorm",

      "mcp__plugin_context7_context7__resolve-library-id",

      "mcp__plugin_context7_context7__query-docs"

    ]

  }

}

```

## Roadmap

**Shipped:**  Multi-model brainstorming & cross-verification • Domain & journal strategy templates • Plot manifest & gap detection • MAGI traceability review • LaTeX math & Gemini fallback chain • Weighted scoring & dynamic personas • Adversarial debate • Murder board & phase gates • Depth-controlled token budget • Session resume (`--resume`) • Artifact contract validation • Standalone phase gates for all sub-skills

**Up next:**

- [ ] Example artifact gallery — real research outputs to showcase the pipeline

- [ ] Terminal demo GIF — one-command walkthrough

- [ ] More domain & journal strategy templates

- [ ] Ubiquitous Context7 — live doc lookups during testing and report writing, not just implementation

- [ ] Conditional variance enforcement — smart error-bar policy that catches missing uncertainty without blocking exploratory runs

- [ ] Cost estimation — token budget preview before execution

## Contributing

Contributions welcome — especially new domain templates. See [CONTRIBUTING.md](CONTRIBUTING.md).

## License

[MIT](LICENSE)