Semiont turns your documents into a knowledge base that people and AI agents build together, with everything it knows pinned to the passage it came from.
Ask about a passage, and it gathers what to read with it: the documents it connects to, similar passages, and what has been noted about them.
Hand it AI work, such as annotating or drafting, and the work becomes a tracked job that any worker can take, yours included.
What comes back is on the record: who asked, who did the work, and what each claim rests on.
Your AI agent can do all of this through the SDK, in any of its languages, to explore on the spot or to build an application that lasts.
Semiont is open source and can run entirely on your own machines.
Read what you add. Documents, PDFs and images, each with everything that has been marked in it.
Mark what matters. Highlight a passage, comment on it, assess it, tag it, or say what it refers to. On a PDF or an image, mark a region.
Let an agent take the first pass. Ask the knowledge base's AI to highlight, comment, assess, tag or find references across a whole document, then keep what is right and delete the rest.
Connect what refers to what. Link a mention to the document about it, or have one written from what the knowledge base already holds. Every document shows what points to it.
Find things. Search by name or by content, and filter by what kind of thing a document is about.
See who did what. Every change to a document is in its history, with who made it, person or agent, and when.
Start with documents, not a schema
Most knowledge systems are no use until someone has spent weeks designing a schema and entering data.
In Semiont you add your documents and start reading.
Ask an agent to find what they mention, keep what is right, and the connections between your documents grow out of that work.
Try it
The Quick Start sets up a knowledge base on your own machine.
Or start from a demo knowledge base, which already has documents in it.
You work in it in the Semiont Browser, in your web browser or as a desktop app.
Documentation
Analyst: work in a knowledge base, annotating, linking, and reviewing what agents propose.
Operator: run a knowledge base, for yourself or for others.
Builder: build an application, a script or an agent on the SDK, in TypeScript, Rust, Python, Go, Kotlin or Java, and customize annotation and generation, down to a worker running your own logic or model.