Your company's corpus
Your agents turn specs, transcripts, and brainstorms into rich, structured docs — canvases, boards, timelines — that compound into one company knowledge base every agent can read over MCP.
Open source · MCP-native · Deploy anywhere
Every surface renders for your team — and is read by your agents over MCP.
One surface
Product docs and architecture, the sales pipeline and runbooks, decisions, onboarding, OKRs, the org — every kind of company knowledge, authored as MDX in your repo and live below. Each surface renders for your team and serializes for your agents.
| Criterion | SQS | Self-host Redis | RabbitMQ |
|---|---|---|---|
| No broker to run | |||
| Exactly-once delivery | |||
| No vendor lock-in | |||
| Ops on-call | None | High | Medium |
Open $1.7M · 19 prospects
Owns sign-up, verification, and the accounts store.
| Person | Domain | Update | Blockers |
|---|---|---|---|
| Maya Chen | Engineering | Shipped the MCP search ranking; reviewing the canvas serializer PR. | — |
| Dev Patel | DevOps | Migrated CI to the new runner; preview deploys are 2x faster. | Waiting on prod secrets rotation |
| Ren Ito | Frontend | Landed the dual-rep table; starting on the org-chart canvas tile. | — |
| John | QA | Wrote smoke suite for the redirect path; all green on staging. | — |
Accent
Accent weak
Ink
One corpus. Humans and agents.
| Signal | Sev | First action |
|---|---|---|
| 5xx > 1% | P1 | Page on-call |
| p99 > 800ms | P2 | Check cart svc |
| Queue depth ↑ | P3 | Scale workers |
Escalation
The people building this era keep circling one idea from different angles — your knowledge has to be readable by machines, not just people. They're naming the problem. superlore is the answer.
Andrej Karpathy@karpathyPosted on XThe hottest new programming language is English.
Jeremy HowardCo-founder, Answer.AI · proposed /llms.txtFrom a blog post or talkLarge language models increasingly rely on website information, but face a critical limitation: the context window is too small to handle most websites in their entirety.
Tobi Lütke@tobiPosted on XI really like the term “context engineering” over prompt engineering. It describes the core skill better: the art of providing all the context for the task to be plausibly solvable by the LLM.
Andrej Karpathy@karpathyPosted on XContext engineering is the delicate art and science of filling the context window with just the right information for the next step.
Harrison ChaseCo-founder & CEO, LangChainFrom a blog post or talkContext engineering is building dynamic systems to provide the right information and tools in the right format such that the LLM can plausibly accomplish the task.
Simon WillisonCreator of Datasette · co-creator of DjangoFrom a blog post or talkMost of the craft of getting good results out of an LLM comes down to managing its context — the text that is part of your current conversation.
Hahnbee LeeCo-founder, MintlifyFrom a blog post or talkAI agents need to know things, and they need to understand your product, your codebase, and your processes. They do this through documentation.
Guillermo RauchCEO, VercelFrom a blog post or talkYou have to think about a web for end users and humans … But you also have to think about a web for agents.
Sam AltmanCEO, OpenAI · on the Big Technology PodcastFrom a blog post or talkEven if you have the world's best personal assistant, they can't remember every word you've ever said in your life. They can't have read every email. They can't have read every document you've ever written. AI is definitely gonna be able to do that.
Aaron LevieCo-founder & CEO, BoxFrom a blog post or talkWhat AI models and agents need is context, and the context that they need to work off is sitting inside your unstructured data.
Arvind JainFounder & CEO, GleanFrom a blog post or talkSo you have to connect the reasoning and generative power of the models with the context inside your company.
Dharmesh ShahCo-founder & CTO, HubSpotFrom a blog post or talkOur systems capture what happened, but not the why.
Mustafa SuleymanCEO, Microsoft AI · on the Big Technology PodcastFrom a blog post or talkIt's going to remember all the big facts about your life … over time, it's going to start to build a richer understanding of who you are and what you care about.
Des TraynorCo-founder, IntercomFrom a blog post or talkIf a bot sees a question, doesn't know the answer, and sends it to a human, it creates a pretty magical experience if the human can answer the question, or train the bot, and update the documentation all in one fell swoop.
Garry Tan@garrytanPosted on XExa is the truth — organizing the world's information but for AI agents everywhere is an extreme unlock for human knowledge.
About 80% of our own docs are read by agents, not people. superlore gives Claude 10× better context than scattered wikis.
Everyone's circling the same idea: your knowledge has to be readable by machines, not just people. That's the whole point of superlore.
Quotes are about the problem space, not endorsements. One card is first-party.
You brainstormed on a whiteboard, closed the tab, and the thinking drifted. Now your AI thinks on the canvas inside the document and shows you — instead of handing back a wall of Markdown to decipher.
superlore turns the wall of Markdown an agent hands back into this live canvas inside your doc — the board your team reads is the typed graph your agents read.
Dual representation
Write one MDX block. It becomes a human surface and a clean, typed node/edge graph — no second authoring step, no drift. The board on the left is the graph on the right.
Release · 2.4
Tighter rate limits at the edge, a cache-first redirect path, and a wider key space. The architecture below is the same typed graph a support agent queries.
Use case · Product docs & releases
A release page with a native architecture diagram and a changelog that's an actual timeline. The diagram an engineer reads is the typed graph a support agent queries — from one MDX source, no second authoring step.
This is a live superlore page, not a screenshot.
Use case · Company knowledge base
Capture how the company actually works: every prospect and where it sits, the service that owns sign-up, the digest from last week's sync. The team scans the pipeline board before every planning call; your agent answers from the same typed graph behind it.
The single screen we scan before every planning call — every prospect, its stage, and the next step.
A live superlore page — the same MDX your team reads and your agent queries.
The loop
The doc is the canvas. Drop comments right on it, Send to agent, and watch the prose and the diagram update together as the thread resolves.
The new standard
Author once; every agent reads the same structured source over MCP — one URL, the typed graph, not a screenshot. And it renders right in the editor your team already uses, so adopting it is a no-brainer.
search(query)query: stringget_page(path)path: stringlist({ kind, tag }){ kind?, tag? }navigate(target)target: stringget_component_data(id)id: stringCompatible with any agent client — Claude, Cursor, Windsurf — over one MCP URL. The typed graph, not a screenshot.
One superlore preview extension runs in VS Code — and Cursor and Windsurf are VS Code-compatible forks, so the same extension renders superlore live there too.
Own it
superlore is open source. Ship it to Vercel, Cloudflare, or your own box — your knowledge stays yours. Scaffold and author it with the superlore skills, preview it in your editor, and deploy when you're ready.
Private and secure when you want it — Google SSO and an org gate flip on from config, and your MCP inherits the same policy.
One corpus. Humans and agents.
Want it fully managed? superlore Cloud hosts your KB and its MCP for you — join the waitlist.
Open source · MCP-native · Auth-ready · Deploy anywhere