Why I Built Atanor: Escaping AI Chat Silos & Bringing Data Portability to LLMs
Hey IH! 👋
Like many of you using AI in your daily development workflow, I heavily rely on context-switching across different LLMs. I’ll start a technical discussion or architecture breakdown with Claude, validate the code logic with ChatGPT, and cross-check reasoning with Gemini or DeepSeek.
The Pain Point: Vendor Lock-In & Broken Formatting
Every time I tried to bring context from one AI to another, standard copy-pasting was a nightmare:
Code blocks lost indentation.
Tables morphed into unreadable text walls.
Nested lists completely broke.
Beyond the formatting frustration, there was a deeper issue: data lock-in. Those technical conversations contained valuable knowledge I built up, yet they were trapped inside proprietary vendor UIs. I couldn't treat my own chats like local, readable documents I actually owned.
The Solution: Atanor
To fix this, I built Atanor — a lightweight Chrome extension that turns AI chat sessions into clean Markdown files with a single click.
Key Technical & Design Decisions:
🧠 Preserves Structure: DOM parsing automatically detects and converts code blocks, tables, lists, and formatting into standard Markdown syntax without mangling.
🔒 100% Local & Privacy-First: Parsing and Markdown generation run entirely client-side in your browser. Zero conversation data is sent to external servers unless you explicitly send it to the editor.
🌐 Multi-Platform: Works seamlessly across ChatGPT, Claude, Gemini, Copilot, DeepSeek, Qwen, MiniMax, Z.ai, and Kimi.
✏️ Integrated Editor: Option to send clean Markdown directly into a built-in browser editor to refine notes before saving.
Building in Public & What's Next
This is v1, and I’m actively building this in public with a primary focus on data portability for AI workflows.
I’d love to get genuine feedback from fellow indie hackers:
Workflow: How do you currently organize or archive your technical AI conversations and research?
Integrations: Which additional LLM web clients or local interfaces (e.g., Ollama UIs) should I prioritize next?
Edge cases: Any specific UI components or formatting elements that break for you during export?
Give it a try at https://atanor.app/en/ — I'd really appreciate your thoughts, feedback, and constructive criticism!
The data-portability point stood out more than the Markdown export itself.
From the people who've tried Atanor so far, are they already treating their AI conversations as something worth preserving and moving between tools, or is that behavior mostly emerging after they see what Atanor makes possible?
To be completely honest, Atanor was born out of scratching my own itch! I juggle multiple LLMs daily and got frustrated with broken formatting when moving context, but I wasn't even sure if others did the same or if I was just an edge-case power user.
Also, because Atanor is 100% local and privacy-first, I intentionally don't collect telemetry or track user behavior.
Since it's super early days, I don't have enough user data yet to tell whether people had this habit beforehand or if the tool unlocked it for them. That's actually one of the main reasons I'm sharing it here: to find out if other devs share this workflow!
How about you? Do you find yourself moving context across models, or do you mostly stick to one?
That’s helpful context. The distinction between solving a personal workflow problem and discovering whether it generalizes is an interesting one.
I’d like to continue the conversation outside the thread. What’s the best email to reach you on?
There is an email address in my profile
Thanks! I’ve just sent it over.
Looking forward to hearing your thoughts whenever you have a chance.