Open-source SEO analysis workspace for audits, AI review, and structured reports
Language: English | Russian
ToraSEO turns technical checks, AI interpretation, and visual reporting into one calm desktop workflow. Start with a URL or article text, choose how the AI should work, and get evidence-backed recommendations you can act on.
ToraSEO is built for people who need more than a raw crawler output, but less chaos than a pile of disconnected SEO tools. It combines:
- a desktop app for analysis setup, progress, reports, and exports
- an MCP server that runs structured checks and sends evidence back to the app
- Claude Desktop and Codex instruction packages for guided external AI workflows
- an in-app
API + AI Chatpath for provider-backed analysis inside ToraSEO
The goal is simple: help you see what is weak, what is strong, and what to fix first.
Prepare the analysis Paste an article, choose the context, and select the checks that matter for the current audit. |
Read the report visually Turn scan evidence into a compact dashboard with readiness, coverage, metrics, and recommendations. |
MCP + Instructions mode Run structured checks through Codex or Claude Desktop. The external AI client calls ToraSEO MCP tools, then the app receives structured results. |
API + AI Chat mode Run the analysis inside ToraSEO with your configured provider and model, then review the findings in the built-in chat and report. |
|
Start from a clear audit path.
ToraSEO helps you choose the right analysis mode before the scan begins, so the report is focused instead of overloaded. |
|
Find weak spots without guessing.
The app separates deterministic scan facts from AI interpretation, making recommendations easier to trust and verify. |
|
Move from findings to next steps.
Reports are designed to answer what matters first: what is broken, what is promising, and what should be improved next. |
|
Build toward stronger search visibility.
ToraSEO is preparing the ground for Tora Rank: a clearer way to compare quality, readiness, and competitive gaps over time. |
The current public direction covers five analysis families:
| Analysis | What it is for |
|---|---|
| Text | Article quality, structure, readability, AI-style signals, SEO packaging, and risk flags |
| Compare two texts | Text A/B comparison, content gaps, similarity risk, style differences, and improvement plans |
| Page by URL | Page text extraction plus article-style checks for a specific URL |
| Site by URL | Technical and on-page audit checks for a single website |
| Site comparison by URL | A compact competitive dashboard for up to three websites |
Some areas are still marked as in development inside the app. The public roadmap is intentionally careful: ToraSEO should show useful evidence first, then layer scoring and ranking logic on top.
Choose the path that matches your workflow.
Best for users who want the visual workspace and either runtime path:
- Download the latest release from GitHub Releases.
- Install the desktop app.
- Choose
MCP + InstructionsorAPI + AI Chaton the home screen. - If you use
API + AI Chat, configure your provider and model in Settings.
Best for users who want the tool layer directly:
git clone https://github.com/Magbusjap/toraseo.git
cd toraseo/mcp
npm install
npm run buildThen register the server in your MCP-compatible client:
{
"mcpServers": {
"toraseo": {
"command": "node",
"args": ["/absolute/path/to/toraseo/mcp/dist/index.js"]
}
}
}Full setup details live in mcp/README.md.
Best for users who want guided audit runs inside an AI client:
| Workflow | Entry point |
|---|---|
| Claude Bridge Instructions | claude-bridge-instructions/README.md |
| Codex Workflow Instructions | toraseo-codex-workflow/README.md |
Download the ZIP assets from the unified Releases page. Do not use auto-generated source archives for installation.
ToraSEO is a multi-surface repository. Several folders are independently useful entry points, so they keep their own README files.
| Surface | Purpose | Entry point |
|---|---|---|
| Root repo | Product overview, release status, documentation map | README.md |
| Desktop app | Native UI, bridge mode, provider settings, reports | app/README.md |
| MCP server | Tool execution layer for scans and bridge data | mcp/README.md |
| Claude Bridge Instructions | Claude-side setup and workflow package | claude-bridge-instructions/README.md |
| Codex Workflow Instructions | Codex-side setup and bridge workflow package | toraseo-codex-workflow/README.md |
| QA docs | Manual checks and smoke-test support | qa/README.md |
ToraSEO is designed so each layer stays independently useful:
- App: status, setup, progress, reports, exports, and native AI chat
- MCP server: scan execution and structured bridge data
- Instruction packages: Claude-side and Codex-side workflow orchestration
Three principles drive the design:
- Evidence first: deterministic scan facts come before model interpretation.
- Composable surfaces: app, MCP, and instruction packages can be used together or separately.
- Explicit trust boundaries: provider secrets, bridge handshakes, and approval flows stay visible.
For deeper design rationale, see docs/ARCHITECTURE.md.
These are current product boundaries, not hidden promises:
- ToraSEO does not replace Search Console, Yandex Webmaster, backlink providers, or full rank trackers.
- It does not claim live SERP visibility, clicks, impressions, or popularity without an official connected source.
- AI-writing probability and AI trace signals are editing heuristics, not proof of authorship.
- Tora Rank is a future direction, not a finished public scoring system in this release.
- Documentation hub
- FAQ
- App README
- MCP README
- Claude Bridge Instructions README
- Codex Workflow Instructions README
- Architecture overview
- Model compatibility notes
- Crawling policy
- Security policy
- Changelog
The fastest ways to help right now:
- Star the repository
- Open an issue with product feedback, bugs, or workflow friction
- Run a real audit and share what worked or broke
- Report security issues privately per SECURITY.md
Formal contribution guidance can expand later, but practical feedback and targeted fixes are already welcome.
Licensed under the Apache License 2.0.
Built by @Magbusjap | Report issue | Security policy | Latest release