Hey Product Hunt 👋 I'm Anirudh, part of the dev team behind Atlas.
Atlas builds your company's context graph: your brand, your voice and how you actually operate, all extracted and connected into one structure. And the whole point is that you own it.
Your company's context need not live inside Claude or OpenAI. With Atlas it's yours: plug the graph into any AI tool your team uses, switch tools tomorrow, and your context comes with you.
Three things we cared about:
1. It builds a real context graph from your brand, voice and processes, connected.
2. Its not locked to any single LLM provider, usable anywhere. You own it.
3. Setup is just plugging in your sources (your website, a few docs). We take care of the extraction. Under 5 minutes.
It's built on the Nanonets document-extraction engine, ranked #1 for document IDP and used by more than a third of the Fortune 500.
We're opening the Founding 200: $99/mo per company, cancel anytime, with white-glove setup where we build your first context with you. For anyone here from Product Hunt today, that white-glove setup is on us. Just drop a comment and I'll reach out.
I'll be in the comments all day. I'd genuinely love your feedback: would owning your company's AI context, instead of re-explaining it to every tool, be useful for your team?
Report
What's different about how context/knowledge graphs work in comparison to Claude skills? If I want someone else's agent to follow my rules, we can just share skills, right?
@aarsh_desai Great question! Unlike Claude’s skills, which are a static, non-evolving set of instructions that can't serve assets, Atlas constantly learns from your conversations. It continuously extracts the latest data to ensure your context graph never goes stale.
Plus, we provide a skill file that plugs our tools directly into your agent. This means you can start building pitch decks, creating brand videos, and pulling metrics immediately in your first chat.
Report
@anirudh_kumar_yadiki Claude's skills do update as well, besides that when you say it learns from conversations, does that mean I need to give Atlas access to my Slack workspace for it to function as intended? Also, will it auto-update the rules it learns from my conversations or will I have the option of controlling what gets added as a rule?
@aarsh_desai Yes, you can give it access to slack but currently you can start using it just mentioning your company url and attaching any reference docs. And yes it does auto update rules that it learns from your conversations, and you can edit or remove them from Atlas Rules Dashboard.
Report
What are the tools you can connect, could you pls specify some?
@vinit_mehta2 Currently the sources of extraction of context would be your company web pages, any reference docs of any file type and your notion. We will build a detailed context graph that all the employees from your company can immediately start using once they connect their agents to Atlas mcp.
Report
Can multiple people edit and contribute to the context?
Hi @vaibhavchhimpa , yes you can have multiple admins for your org, and they can edit/delete company wide rules and context , and each member admin and non admin have their own personal context so their personal preferences also get saved and reused.
The insight that company context should live outside individual AI tools rather than being re-input in each system prompt is sound. Most teams end up maintaining duplicate context blobs in Cursor, Claude, and internal tools that drift out of sync. How does Atlas push updates to connected tools when company guidelines change? Is it pull-based querying or active propagation to each integration?
@anand_thakkar1 we make sure the connected tools always fetch the latest fresh context of the company through mcp connectors.
Report
AI can extract data really well, but trust is a different challenge. At what point do your customers stop double-checking the output and start relying on it confidently?
thanks @harini_mukesh , we strongly believe that once agents start using context graphs as their memory source, customers of those agents will stop double checking the output
The premise is right. The failure mode with most AI tools isn't the model, it's that every session starts cold and you end up re-explaining your positioning, your audience, your tone, your internal terminology, over and over across a dozen different tools.
What I'm curious about is how Atlas actually propagates that context. Is there a central knowledge layer that each connected tool reads from, or are you syncing context into each tool's own memory or system prompt? And when your company context changes, say you rebrand or shift positioning, how does that update flow through to the tools that already have the old version baked in?
@fberrez1 currently, we maintain a central knowledge layer (context graph) containing both org specific and personal use specific rules and assets. And Atlas does sync with its sources frequently ( your web pages, notion , slack) so that the context doesn't go stale.
@sravanth_talluri any person in your company can just signup and setup the mcp tool in ai agent of your choice (max two steps) and then your entire org can start using the company context .
Report
Company context is only valuable if the agent can show where each answer came from. The hard operational problem is not just memory, it is provenance, stale-source handling, and knowing when to ask before acting.
yes @krekeltronics , this is a problem Atlas looks to solve by making sure we sync regularly with the connected sources, and making sure we extract processes and not just knowledge so that the ai agent using our context takes the correct decisions.
Replies
Nanonets
Hey Product Hunt 👋 I'm Anirudh, part of the dev team behind Atlas.
Atlas builds your company's context graph: your brand, your voice and how you actually operate, all extracted and connected into one structure. And the whole point is that you own it.
Your company's context need not live inside Claude or OpenAI. With Atlas it's yours: plug the graph into any AI tool your team uses, switch tools tomorrow, and your context comes with you.
Three things we cared about:
1. It builds a real context graph from your brand, voice and processes, connected.
2. Its not locked to any single LLM provider, usable anywhere. You own it.
3. Setup is just plugging in your sources (your website, a few docs). We take care of the extraction. Under 5 minutes.
It's built on the Nanonets document-extraction engine, ranked #1 for document IDP and used by more than a third of the Fortune 500.
We're opening the Founding 200: $99/mo per company, cancel anytime, with white-glove setup where we build your first context with you. For anyone here from Product Hunt today, that white-glove setup is on us. Just drop a comment and I'll reach out.
I'll be in the comments all day. I'd genuinely love your feedback: would owning your company's AI context, instead of re-explaining it to every tool, be useful for your team?
What's different about how context/knowledge graphs work in comparison to Claude skills? If I want someone else's agent to follow my rules, we can just share skills, right?
Nanonets
@aarsh_desai Great question! Unlike Claude’s skills, which are a static, non-evolving set of instructions that can't serve assets, Atlas constantly learns from your conversations. It continuously extracts the latest data to ensure your context graph never goes stale.
Plus, we provide a skill file that plugs our tools directly into your agent. This means you can start building pitch decks, creating brand videos, and pulling metrics immediately in your first chat.
@anirudh_kumar_yadiki Claude's skills do update as well, besides that when you say it learns from conversations, does that mean I need to give Atlas access to my Slack workspace for it to function as intended? Also, will it auto-update the rules it learns from my conversations or will I have the option of controlling what gets added as a rule?
Nanonets
@aarsh_desai Yes, you can give it access to slack but currently you can start using it just mentioning your company url and attaching any reference docs.
And yes it does auto update rules that it learns from your conversations, and you can edit or remove them from Atlas Rules Dashboard.
What are the tools you can connect, could you pls specify some?
Nanonets
@vinit_mehta2 Currently the sources of extraction of context would be your company web pages, any reference docs of any file type and your notion. We will build a detailed context graph that all the employees from your company can immediately start using once they connect their agents to Atlas mcp.
Nanonets
Hi @vaibhavchhimpa , yes you can have multiple admins for your org, and they can edit/delete company wide rules and context , and each member admin and non admin have their own personal context so their personal preferences also get saved and reused.
FastCut
can this also work with claude code?
Nanonets
@shrish_dwivedi yes, we support claude code
The insight that company context should live outside individual AI tools rather than being re-input in each system prompt is sound. Most teams end up maintaining duplicate context blobs in Cursor, Claude, and internal tools that drift out of sync. How does Atlas push updates to connected tools when company guidelines change? Is it pull-based querying or active propagation to each integration?
Nanonets
@anand_thakkar1 we make sure the connected tools always fetch the latest fresh context of the company through mcp connectors.
AI can extract data really well, but trust is a different challenge. At what point do your customers stop double-checking the output and start relying on it confidently?
Nanonets
thanks @harini_mukesh , we strongly believe that once agents start using context graphs as their memory source, customers of those agents will stop double checking the output
Foyer
The premise is right. The failure mode with most AI tools isn't the model, it's that every session starts cold and you end up re-explaining your positioning, your audience, your tone, your internal terminology, over and over across a dozen different tools.
What I'm curious about is how Atlas actually propagates that context. Is there a central knowledge layer that each connected tool reads from, or are you syncing context into each tool's own memory or system prompt? And when your company context changes, say you rebrand or shift positioning, how does that update flow through to the tools that already have the old version baked in?
Nanonets
@fberrez1 currently, we maintain a central knowledge layer (context graph) containing both org specific and personal use specific rules and assets. And Atlas does sync with its sources frequently ( your web pages, notion , slack) so that the context doesn't go stale.
What's the best way to get started for a company?
Nanonets
@sravanth_talluri any person in your company can just signup and setup the mcp tool in ai agent of your choice (max two steps) and then your entire org can start using the company context .
Company context is only valuable if the agent can show where each answer came from. The hard operational problem is not just memory, it is provenance, stale-source handling, and knowing when to ask before acting.
Nanonets
yes @krekeltronics , this is a problem Atlas looks to solve by making sure we sync regularly with the connected sources, and making sure we extract processes and not just knowledge so that the ai agent using our context takes the correct decisions.