Praxos – Context Management for AI Agents
Hey HN! We're Lucas and Soheil, the founders of Praxos (https://praxos.ai). Praxos is a context manager for AI Agents, providing everything you need to build stateful agents that don't break in production. Praxos can parse any data source, from unstructured PDFs and API streams to conversational messages, to structured databases, and transform them into a single Knowledge Graph. Everything in this graph is semantically typed and its relationships are made explicit, turning data into a clean, queryable universe of understanding that AI can use without making mistakes. Whether you…
In plain words
Praxos is a context management platform for AI agents that ingests data from multiple sources—including PDFs, APIs, databases, and conversations—and transforms it into a unified Knowledge Graph with explicit semantic relationships. The platform enables developers to build stateful agents that maintain consistent performance in production by providing clean, queryable data structures that reduce AI errors. It supports complex reasoning tasks across single and multi-step chains without requiring repeated data queries.
written from the facts on this page · September 2026
From the sources
In the maker’s words, at launch
Hey HN! We're Lucas and Soheil, the founders of Praxos (https://praxos.ai). Praxos is a context manager for AI Agents, providing everything you need to build stateful agents that don't break in production. Praxos can parse any data source, from unstructured PDFs and API streams to conversational messages, to structured databases, and transform them into a single Knowledge Graph. Everything in this graph is semantically typed and its relationships are made explicit, turning data into a clean, queryable universe of understanding that AI can use without making mistakes. Whether you need to query for the answer to a question or to extract data in a way that makes sense for the current use case, Praxos does it all, with no requerying needed. This enables AI apps to parse data end-to-end, and then act on it to deliver outputs across single-chain and multi-chain reasoning steps. Intermediate, final, and user-edited outputs can be added back to the knowledge graph, allowing Praxos to learn on the fly. When we were building in insurance, we often ran into two major problems deploying AI: First, LLMs would prove incapable of parsing documents such as property schedules and insurance policies. For reference, a property schedule may be a 50-page collection of Word, Excel, and PDF documents detailing construction, usage, and geographical information about a collection of physical properties. Recreating one object (a property) would mean combing through the files establish semantic, conceptual, spatial, and sometimes implicit linkages between the data. The outcome: relationship information would be lost, left blank, or hallucinated. Second, repeated calls to search, retrieve, and update information would sometimes lead to cascading errors. This became more frequent across complex tasks such as reading a document, fetching previous user information, performing a calculation, storing it, and then presenting it to the user. We realized that for AI to deliver more useful and accurate responses that correctly use relationships in the document, these relationships need to be made explicit. Much of the contextual information is represented without the usage of words. In turn, this means that we cannot directly interact with them programmatically, and LLMs are forced to interpret them themselves, every single time. That’s when we started building Praxos. We've set up a self-serve option with a free tier (up to a data cap) for hobbyists and early-adopters. For context (no pun intended), this should cover you for up to 200 document pages. You can register here: https://data.praxos.ai/sign-up. Our first version is an SDK meant to cover you across all your data extraction, retrieval, and update needs. Here's how it works: Organizing information: Praxos sorts information into ontologies, which are structured schemas for storing data. These allow you to introduce predefined types, attributes, and relationships that guide how the knowledge graph is built and interpreted. Processing input data: Praxos can handle any data source, ranging from PDFs to tabular data, JSONs, and dialog-like exchanges. Extraction is performed end-to-end. You don't need to OCR, chunk, or pre-process your inputs. Processing is as simple as passing in your file and selecting an ontology. Retrieving information / memories: For each query, Praxos searches and retrieves related stored information by leveraging a combination of graph traversal techniques, vector similarity and key-value lookups. Search objects will return both the entities/their connections, as well as a sentence. We’d love to hear what you think! Please feel free to dive in, and share any thoughts or suggestions with us over Discord (https://discord.gg/wVmrXD2dJA). Your feedback will help shape where we take Praxos from here!
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 26d ago · cactuscompute.com
Launched alongside, July 2025
the whole month →
Trickle - Magic Canvas ▲1,064The 1st Agentic Canvas for building apps visually with AI
AI · 2025 · trickle.so
- TB
Life & fun · 2025 · trytender.app
- TY
Today marks ten years, 3653 consecutive days, of running at least one mile every day under the USRSA rules [1]. To celebrate, I built an interactive dashboard that turns a decade of GPX files into charts you can explore. Running has truly changed my life: I've made lifelong friends, explored beautiful places, and more importantly invested into my own health and fitness, which I'm starting to see the positive benefits as I get older. The stack is pretty simple: a NextJS app, with a Postgres database to keep all my running data, and all the stats are pre-computed and cached in Redis, so I…
Dev tools · 2025 · nodaysoff.run
- DA
Made this website as an exercise in vibe-coding and GCP. It was posted about a few times around the internet, on sites like Morning Brew (https://www.morningbrew.com/issues/business-buzzkill), MetaFilter (https://www.metafilter.com/209703/Draw-A-Fish), boingboing.net, etc. I think it's cute! I built a basic CNN trained against penises and swastikas, and then anything that doesn't hit the 63% confidence score gets sent to a mod queue, a vibe-coded fish-tinder…
Dev tools · 2025 · drawafish.com
