PaperDebugger
An AI-powered academic writing assistant that helps you debug and improve research papers directly in Overleaf.
Evidence shows extension claims read-only, but no permission list or least-privilege analysis; user confirmation (e.g., before insertion) not documented; data flow (e.g., to OpenAI) not detailed; sensitive data handling (e.g., API keys) not described; dependency security lacks vulnerability scanning or audit; external effects (e.g., modifying Overleaf project) claimed read-only but unverified; rollback mechanism not mentioned; source attribution (e.g., citations) not explained. Deductions: lack of implementation details and verification.
Self-consistency: README and architecture description consistent, but no error handling details; dependency availability: dependency list complete, but no version compatibility notes; failure messages: no examples of error handling or user prompts. Deductions: documentation does not cover error scenarios.
Audience clear (researchers), scenarios clear (academic writing); capability boundaries (read-only) stated but not detailed; trigger precision (e.g., when suggestions are triggered) not specified; environment fit (Overleaf integration) described but no configuration details. Deductions: lack of specific configuration and boundary examples.
Information architecture clear (README has TOC); install notes present (Chrome store and self-host); naming stability not stated; examples and FAQ missing; known limitations not listed; license clear (AGPL-3.0); versioning and changelog not provided; maintenance responsibility not clear. Deductions: missing examples, limitations, and version info.
Output usability: one-click insertion and comments, but output format not described; marginal value: multi-agent orchestration has potential, but no comparison; cost-benefit: no pricing or resource consumption info. Deductions: lack of concrete effectiveness data.
Claims (e.g., multi-agent) have architectural support, but no tests or verification; cross-source verification not provided; facts and inferences not separated. Deductions: lack of verifiable evidence.
- Extension claims read-only, but no permission list; users should review actual permissions.
- Data flow is opaque; users should know what is sent to AI services.
- Dependency security not audited; users should check for vulnerabilities.
- Rollback mechanism not mentioned; users should back up projects.
What does this agent do, and when should you use it?
PaperDebugger is a plugin-based multi-agent system for academic writing, review, and editing. It runs as a Chrome extension that integrates with Overleaf, allowing researchers to receive intelligent suggestions without leaving the editor. The system is powered by XtraMCP, a custom MCP-based orchestration engine that simulates the full academic workflow: Research → Critique → Revision, enabling multi-step reasoning, reviewer-style critiques, and structured revision passes. The backend is built with Go, using Gin and gRPC, MongoDB for storage, and integrates with the OpenAI API. The extension is read-only, ensuring it never modifies the user's project, and offers one-click insertion of AI responses, a comment system, a prompt library, and privacy-first design.
After installing the Chrome extension and opening an Overleaf project, users can click the PaperDebugger icon on the top-left to open a chat interface. It leverages XtraMCP multi-agent orchestration to perform literature-grounded research, AI-conference review, citation verification, and domain-specific revision. Users can insert AI responses directly into their project with one click or generate comments. The extension only reads project content, ensuring security. The backend supports custom endpoint configuration, allowing self-hosting, but requires Overleaf login for self-hosted backends.
- Researchers writing LaTeX papers in Overleaf who want real-time AI feedback and suggestions for improvement.
- Authors who need reviewer-style critique to enhance the quality of their manuscripts, using the AI-conference review feature.
- Users who want to verify citations in their papers for accuracy through the citation verification capability.
- Writers who prefer to discuss paper structure via a chat interface without leaving the editor.
- Teams that require multi-step reasoning and structured revision passes to ensure rigorous academic writing, using the Research → Critique → Revision workflow.
- Users who want to use custom prompt templates for tailored assistance in different writing scenarios.
What are this agent's strengths and limitations?
- Deep integration with Overleaf, eliminating the need to switch editors for AI assistance.
- Multi-agent orchestration supports a complete academic workflow (research, critique, revision), exceeding standard chat assistance.
- Privacy-first: read-only access ensures user content is never modified.
- Prompt library and one-click insertion streamline the writing process.
- Limited to Chrome and Overleaf; no support for other editors.
- AI functionality depends on OpenAI API, requiring network and possible costs.
- Self-hosting requires additional configuration and Overleaf login for authentication.
- Read-only nature means users must manually apply suggestions, adding extra steps.
How do you install or deploy this agent?
- Install the extension from the Chrome Web Store: https://chromewebstore.google.com/detail/paperdebugger/dfkedikhakpapbfcnbpmfhpklndgiaog
- Once installed, open any Overleaf project.
- Click the PaperDebugger icon on the top-left to begin.
Note: For self-hosted backends, go to Settings, click the version number 5 times to enable Developer Tools, then configure the backend endpoint.
How do you use this agent?
- Install the extension, then open an Overleaf project.
- Click the PaperDebugger icon to open the sidebar.
- Type questions or requests in the chat box, e.g., 'Please review my introduction.'
- After receiving suggestions, use the 'Insert' button to add AI responses to your document, or use the 'Comment' feature to generate comments.
- Access the Prompt Library for preset templates or create custom prompts.
- For self-hosting, configure a custom endpoint and refresh the page.