Recurse turns raw input into a living, semantically typed knowledge graph so you and your AI systems can explore, connect, and reason — not just retrieve.
Recurse turns raw input into structured, actionable context
Add any type of content and we transform it into a living, semantically typed knowledge graph that you (and your AI agents) can act on, reason through, and build on top of.
Zero config needed. Change just one line of code to make your AI context-aware.
Not quite.
It’s built a bit differently...
Recurse is memory infrastructure for systems that actually understand.
Most context management systems are optimized for retrieval, not exploration.
Ask a question, get back what looks most similar to your query. This works if you know what you're looking for. But it systematically prevents the kind of exploration that leads to genuine understanding and novel insights.
Standard retrieval systems don't allow you to discover connections you didn't know existed, stumble onto relevant context from unexpected sources or follow threads that diverge from your initial question.
Recurse favors depth over similarity, relationships over rankings and evolution over static storage.
Learns your structure automatically
Drop in any content—research papers, meeting notes, code documentation—and watch structure emerge. Recurse learns the patterns in your material and creates meaningful relationships automatically. No schemas to define, no categories to maintain.
| Adapts to any content type |
| Zero configuration needed |
| Learns your domain automatically |
Assembles context bundles that enrich your inquiries
When you need information, Recurse assembles complete context bundles—not scattered chunks. Copy with one click for any AI tool, or connect your agents directly. Context arrives pre-assembled with sources cited and relationships preserved.
| One-click context export |
| Direct agent injection |
| Sources always cited |
Lets you navigate semantic structure
Follow how ideas actually connect. Arguments link to evidence, decisions trace back to discussions. Both you and your AI agents can navigate these reasoning chains, discovering relationships that similarity search would miss.
| Typed relationship traversal |
| Agent-accessible navigation |
| Multi-hop exploration |
Evolves your knowledge base over time
Your knowledge graph isn't static—it grows and refines as you add content. Recurse preserves complete history with timestamps and explanations of what changed, so you can trace how understanding developed.
| Complete version history |
| Track knowledge evolution |
| Query past states |
Keeps your knowledge base always up to date
Subscribe to RSS feeds, documentation sites, and newsletters. New content flows in automatically, gets processed, and connects to your existing knowledge. Stay current without manual uploads.
| Automatic ingestion |
| Monitor any feed or URL |
| Continuous updates |
Lets you tap into expert knowledge
Subscribe to domain experts' curated knowledge bases directly—their sources, synthesis, and connections become part of your context. Your AI draws from their expertise automatically, updating as they publish.
| Access expert knowledge |
| Share curated collections |
| Automatic updates |
Navigate your knowledge conversationally
Upload & Chat
Drop PDFs, text files, or paste URLs. Then ask questions naturally—answers come grounded in your sources. Works like ChatGPT but connected to your knowledge.
Assemble Context
Build context bundles with one click. Select what matters, export for any AI tool, or inject directly into your agent workflows. Sources cited, relationships intact.
Explore Visually
See how your ideas connect. Navigate relationships between concepts, discover patterns, and trace reasoning chains through your knowledge.
For Developers
Stop reinventing memory infrastructure. Integrate Recurse with one line of code. Works with any AI provider.
Start Exploring
Access the chat interface, graph visualizer, and source management tools. No coding required.
Our interface makes your knowledge immediately explorable. Upload documents, ask questions in natural language, trace connections through your graph. And if you need programmatic access for building applications or custom integrations on top of your knowledge base, there's a full API underneath.
AI Assistants
that retain context across conversations and sources, trace reasoning chains, and build on accumulated understanding
Knowledge Workspaces
that adapt to your thought patterns, surface unexpected connections, and evolve as you learn
Research Tools
that navigate relationships between papers, track how understanding evolved, and discover cross-domain insights
Collaborative Agents
that share structured knowledge, reason through complex decisions, and maintain context across teams
Make your AI context-aware with just one line of code
import OpenAI from 'openai'; const client = new OpenAI({apiKey: process.env.OPENAI_API_KEY,baseURL: 'https://api.openai.com/v1/',defaultHeaders: { 'X-API-Key': process.env.RECURSE_API_KEY, 'X-Recurse-Scope': 'my_project' } });// Use the client normally const completion = await client.chat...
Automatic context injection
No need for manual context engineering. Use our proxy to get started in minutes.
Swap your base URL to point to the Recurse proxy. When you send a request to your AI provider, Recurse retrieves context from your knowledge graph and injects it into the request. When your model responds, Recurse extracts semantic frames and stores them back. Your code sees a standard response. Enrichment and extraction happen automatically.
How Recurse Compares
Recurse goes beyond traditional RAG and Graph RAG by creating living knowledge structures that adapt and evolve with your content.
Full comparison| Feature | Traditional RAG | Graph RAG | Recurse |
|---|---|---|---|
| Understanding | Similarity matching | Entity extraction | Semantic structure |
| Navigation | Flat chunk retrieval | Entity relationships | Recursive traversal |
| Adaptation | Static | Fixed ontology | Emergent schemas |
| Memory | Session-based | Static graph | Living substrate |
Join our beta
We are looking for teams that want to put our approach to the test
Are you building AI assistants, managing extensive knowledge bases, or streamlining research workflows?
Tell us a bit about your use case and join our beta. We are currently onboarding new users on a bring your own key basis (learn more).
Context isn’t optional anymore
Whether you’re designing agents, research platforms, or adaptive interfaces — Recurse gives you the structure and semantic depth traditional tools lack. It’s the memory substrate for AI-native infrastructure.