Your AI doesn't know
what you know.

Relatch reads your Markdown docs and builds structured skill files that Claude and OpenAI Codex can actually reason from. Not raw pasted text. Persistent, typed intelligence.

Start for Free See How It Works
document_042.pdf

Drop any document

PDF • CSV • XML • DOCX • MD • TXT • JSON

Extracting Structure
Raw Text
Structured
AI Enrichment
Tables
Logic
Rules
Context
SKILL.md
or
agent.yaml + skill.md
name: "usfinance" domain: "financial datasets" content_type: "Model datasets financial knoweldge skill"
├─ Calculate Net Cash Flow ├─ If Positive: Evaluate Capital └─ If Negative: Audit Transfers
AI Ready
The Problem

Your knowledge
is locked away.

You've spent years building expertise in documents, notes, and files. Every AI session starts from zero. Your AI can't reason from raw files; it needs structured skill layers.

Step 1 — Upload

Drop anything.
Every format handled.

PDF, DOCX, CSV, XML, Markdown, code. Relatch accepts your full knowledge stack and routes each file to the right processing path, automatically.

Step 2 — Extract

Every signal,
captured precisely.

Smart OCR + structural parsing pulls meaning from messy files, including tables, lists, domain language, and code patterns. Scan-based PDFs auto-escalate to Mistral OCR. Nothing important is lost.

Step 3 — Enrich

AI detects the
knowledge type.

Relatch identifies whether your document holds persona patterns, codebase logic, process workflows, or domain knowledge, then applies the matching enrichment template. Output is scored before delivery.

Step 4 — Export

Your AI gets
a real skill file.

Download a structured .md for Claude or a .zip for OpenAI Codex CLI. Install once; your AI now has persistent, structured access to everything in those documents.

Adaptive Intelligence

One upload. The right template.

Relatch detects what your document actually contains and automatically builds the correct skill structure. No guessing. No manual selection.

Relatch AI Engine
Persona & Voice
Voice + Style
Codebase Intel
Architecture
Process & Workflow
Workflow Logic
Prof. Domain
Domain Expertise
Financial Data
Data + Schema
Execute
Direct Action
Expertise
Human-in-Loop
Specialist
Scoped Role
Under the Hood

Engineered for how AI actually reasons

Smart OCR routing for scanned docs
pdfjs extracts first. If the PDF is scan-based or diagram-heavy, Relatch auto-escalates to Mistral OCR with OCR.space as fallback. Scanned documents just work.
Adaptive templates + domain knowledge
Relatch detects document type and applies the matching enrichment template. Output structure reflects what the document actually is, not just what it says.
Auto-categorization with relevancy levels
Label uploads as personality, knowledge, instructions, examples, context, or preferences. Each category sends a different focus directive to Gemini.
Easily injectable in Claude and Codex
Export a .md for Claude or a .zip for Codex CLI. One-click install. Your AI has persistent access to your entire knowledge base from that session forward.
100% quality-gated skill files
Output is scored before delivery. Gemini Flash Lite generates first — if it doesn't meet quality threshold, Relatch escalates to Gemini 2.5 Flash. You get the best result, not the fastest.
Private by design
Documents are processed through Relatch's own proxy and immediately discarded. Nothing stored. Nothing used for training. Clerk auth is enforced on every API call — your IP stays yours.
Platform Support

Built for Claude.
Built for Codex.

Two different AI systems, two different skill file formats. Relatch handles both — natively.

Claude by Anthropic
Relatch exports a single structured .md file that installs directly into Claude Projects as a skill. Your AI now has persistent, domain-specific context on every conversation.
Exports as SKILL.md — drop into Claude Projects
5 adaptive templates auto-selected per document type
YAML frontmatter + decision logic + guardrail rules
Quality-scored before delivery via Gemini cascade
Try with Claude
OpenAI Codex CLI
Relatch exports a .zip containing SKILL.md + agents/openai.yaml — the exact format Codex CLI expects. Install once, reason forever.
Exports as .zip with SKILL.md + agents/openai.yaml
3 Codex shapes: EXECUTE, EXPLORE, DEBUG
Structured for Codex CLI's skill-loading protocol
Code-specific extraction preserving function signatures
Try with Codex
150+ AI practitioners
building with Relatch

Developers and knowledge workers turning their documents into AI-ready skills — across the United States, India, Vietnam, Morocco, Ireland, Romania, Australia, and France.

Common Questions

You Ask? We answer.

Relatch is a smart adaptive engine that reads your document, understands what kind of knowledge it contains, and turns it into a structured skill file. It removes the friction of figuring out format, structure, and export logic from scratch.

Because the hard part is not writing the file. The hard part is shaping messy real world information into the right structure in the first place. Relatch does that thinking for you, so you start with a clean, agent ready output instead of a blank page.

Relatch works with PDFs, docs, notes, markdown, code files, spreadsheets, and mixed knowledge files. It analyzes the content, identifies the type of knowledge it contains, and automatically selects the right template, structure, and export format for the output.

Not at all. Relatch is designed for non technical users, researchers, writers, analysts, and developers alike. Upload your files, choose your goal, and Relatch handles the heavy lifting. The generated outputs are ready to use with minimal setup.

Your file is used to generate the structured output you need. The experience is designed to feel simple. Upload the source, Relatch extracts the useful structure, and you receive a skill ready result without manually rebuilding everything yourself.

No. Relatch is designed to remove setup friction, not add it. You do not need Claude and Codex Pro or any complicated configuration to use the product and generate a structured result.

Yes. Relatch is designed with file safety as a core principle. It does not have a backend for storing user files — the moment you refresh your browser your session restarts again. Your session is protected through Clerk so your IP and session access remain protected throughout the process.

Yes. Relatch is currently free during the product validation phase. To keep the service sustainable, each user has daily and weekly usage limits that automatically refresh over time.

Legal

Privacy Policy

Last updated: April 2026

1. Overview

Relatch is a document intelligence tool. This Privacy Policy explains what data we collect, why we collect it, and how we handle it. We have built Relatch with a minimal data footprint by design. Authenticated users can upload documents which are processed to generate structured outputs intended for use as AI context. We aim to keep our data handling lightweight, reasonable, and transparent.

By using Relatch at app.relatch.online, you agree to the practices described in this policy.

2. What We Collect

2.1 Analytics and Behavioural Data

Relatch uses the following third-party analytics tools to understand how users interact with the product. This data is aggregated and does not personally identify you.

You can opt out of Google Analytics by installing the Google Analytics Opt-out Browser Add-on. You can opt out of PostHog by disabling cookies in your browser settings.

2.2 Account Information

To upload and process documents, you need an account. Authentication and session management are handled by Clerk. You can sign in using Google OAuth or GitHub OAuth. When you sign in through one of these providers, Relatch receives a basic identity profile (such as your name, email address, and provider user ID) so that we can associate your session with your account. We do not receive your OAuth password.

We use this account information solely to authenticate you, scope your uploads and outputs to your account, and provide essential account-related communication. We do not sell or rent account data.

2.3 Waitlist and Contact Submissions

Relatch uses Formspree for some contact and waitlist flows on the landing page. If you choose to submit your email address or a message through one of these forms, that submission is delivered via Formspree and used solely to follow up with you about Relatch. We will not send marketing emails unrelated to Relatch. You can request removal of your email at any time by contacting us.

2.4 Uploaded Documents

Documents you upload (PDF, CSV, XML, DOCX, scanned images, and similar formats) are processed client-side where possible. When server-side processing is required, your content is transmitted to third-party AI models solely for extraction and enrichment. We do not retain your uploaded files on our servers beyond what is reasonably needed to deliver the resulting output. We do not use your documents to train any AI model. You are responsible for the content you upload.

3. How Your Documents Are Processed

Relatch uses a multi-stage processing pipeline to extract and enrich content from your documents.

3.1 Text Extraction

3.2 AI Enrichment

After extraction, the text is sent to an AI model for enrichment and skill file generation:

Your document content is transmitted to Google's AI services for enrichment. Please review Google's Privacy Policy for details.

3.3 Smart Analyzer

Before any extraction or enrichment begins, Relatch's smart analyzer scans your upload for corrupted files, unsupported formats, and irrelevant document types. Files that do not pass this check are discarded immediately and never transmitted to any external service.

4. Cookies and Session Storage

Relatch uses cookies and similar browser storage for two purposes: (a) analytics, placed by Google Analytics and PostHog as described above; and (b) authentication, where Clerk uses cookies and session storage to keep you signed in and to maintain your session securely. Authentication storage is essential to keeping accounts functional. Analytics cookies are non-essential and you can manage or disable them through your browser settings.

5. Third-Party Services

The following third-party services may process your data as part of using Relatch:

Each of these services operates under its own privacy policy. We encourage you to review them independently.

6. Data Retention

We retain account information for as long as your account is active. Uploaded documents are not kept beyond what is reasonably needed to deliver the resulting output. Analytics data is retained by Google Analytics and PostHog under their respective retention policies. Waitlist and contact-form submissions are retained by Formspree until you request removal.

7. Your Rights

You can request deletion of your account, your waitlist email, or any contact submission at any time by emailing us at relatchh@gmail.com. We will action reasonable requests within a reasonable timeframe.

8. Children

Relatch does not target children and does not knowingly collect personal information from minors. If you believe a minor has submitted their email address via the waitlist form, contact us and we will remove it promptly.

9. Governing Law

This Privacy Policy is governed by the laws of India. Any disputes arising in connection with this policy shall be subject to the jurisdiction of the competent courts in India.

10. Changes to This Policy

We may update this Privacy Policy as Relatch evolves. When we do, we will update the date at the top of this page. Continued use of Relatch after any changes constitutes acceptance of the updated policy.

11. Contact

For any privacy-related questions, contact us at relatchh@gmail.com.

Terms of Service Back to Relatch
Legal

Terms of Service

Last updated: April 2026

1. Acceptance of Terms

By accessing or using Relatch at app.relatch.online, you agree to be bound by these Terms of Service. If you do not agree, do not use Relatch. These terms apply to all users of the service.

2. Description of Service

Relatch is a document intelligence tool. Authenticated users can upload files including PDFs, CSVs, XML files, DOCX files, scanned documents, and similar formats. Relatch may process those uploads to extract their content, enrich it using AI models, and produce structured AI context outputs (such as skill files) intended for use with AI coding and reasoning agents.

Relatch is currently in active development. Features, availability, and processing pipelines may change without prior notice. We do not guarantee uninterrupted availability of the service.

2.1 Accounts and Authentication

To upload and process documents you need a Relatch account. Authentication is provided through Clerk, and you can sign in using Google OAuth or GitHub OAuth. Your use of those third-party authentication providers is also governed by their respective terms and privacy policies. You are responsible for maintaining the confidentiality of any credentials used to access your account and for any activity that occurs under it.

3. Free Access

Relatch is free to use. There are no subscription fees, usage limits, or paid tiers at this time. We reserve the right to introduce pricing in the future. If we do, we will notify waitlist subscribers and provide adequate notice before any free tier is modified.

4. Acceptable Use and User Content

You are solely responsible for any content, files, or documents you upload to Relatch. You agree not to use Relatch to:

We may suspend or terminate accounts that abuse or misuse the service, violate these terms, or upload content that falls into any of the categories above. Where reasonable, we will try to give notice, but we may act immediately when needed to protect Relatch, other users, or third parties.

5. Document Processing and AI Models

Relatch processes uploaded documents through a multi-stage pipeline. Text extraction is performed using a PDF.js parser, Mistral OCR, or OCR Cloud Space depending on the document type. Enrichment is performed using Google's Gemini AI models. A fallback protocol using internal templates is activated if external AI services are unavailable.

Relatch includes a smart analyzer that screens uploads before processing begins. Corrupted files, unsupported formats, and documents that do not pass the analyzer's checks are discarded automatically and never transmitted to any external service.

Relatch makes no guarantee that outputs generated by the AI enrichment pipeline are accurate, complete, or suitable for any particular purpose. You are responsible for reviewing all outputs before use.

6. Intellectual Property

6.1 Your Content

You retain full ownership of any documents you upload to Relatch. By uploading a document, you grant Relatch a limited, temporary licence to process that document solely for the purpose of generating your requested output. This licence ends when your session ends.

6.2 Generated Outputs

The structured skill files and outputs generated by Relatch from your documents are yours to use. Relatch does not claim ownership over outputs derived from your content.

6.3 Relatch's Property

The Relatch name, logo, interface design, processing logic, internal templates, domain knowledge rules, and all related intellectual property remain the exclusive property of the Relatch team. Nothing in these terms grants you any rights to these assets.

7. Third-Party Services

Relatch integrates with third-party services including Clerk (authentication), Google OAuth and GitHub OAuth (sign-in), Google Gemini, Mistral OCR, OCR Cloud Space, Formspree, Google Analytics, Google Tag Manager, PostHog, and Vercel (hosting). Your use of Relatch is subject to the terms and privacy policies of these services where applicable. Relatch is not responsible for the availability, accuracy, or conduct of any third-party service.

8. Disclaimers

Relatch is provided on an "as is" and "as available" basis without warranties of any kind, express or implied. We do not warrant that Relatch will be uninterrupted, error-free, or that outputs will meet your requirements or expectations.

AI-generated outputs can be incomplete, incorrect, or contextually inappropriate. You bear full responsibility for how you use any output produced by Relatch.

9. Limitation of Liability

To the fullest extent permitted by applicable law, Relatch and its developers shall not be liable for any direct, indirect, incidental, consequential, or special damages arising from your use of the service, including but not limited to loss of data, loss of business, or damages resulting from reliance on AI-generated outputs.

10. Availability and Changes

Relatch is under active development. We may modify, suspend, or discontinue any part of the service at any time without notice. We are not liable to you or any third party for any modification, suspension, or discontinuation of the service.

11. Waitlist and Communications

If you submit your email address through the Relatch waitlist form, you consent to receiving a notification when Relatch launches at full scale. You will not receive unsolicited marketing. You can request removal from the waitlist at any time by emailing relatchh@gmail.com.

12. Governing Law and Jurisdiction

These Terms of Service are governed by the laws of India. Any disputes arising from or in connection with these terms shall be subject to the exclusive jurisdiction of the competent courts in India.

13. Changes to These Terms

We may update these Terms of Service as Relatch evolves. When we do, we will update the date at the top of this page. Your continued use of Relatch after any update constitutes acceptance of the revised terms.

14. Contact

For any questions regarding these terms, contact us at relatchh@gmail.com.

Privacy Policy Back to Relatch