Context & Agents
Context & Agents
See the agent packs that draft your work, and manage the code and documents Next and its agents read.
Next does its best work when it knows your organisation: your products, your policies and the way your software is built. The Context & Agents section of the sidebar is where that knowledge lives. It has three pages:
- Agent Packs – the teams of AI agents that produce drafts, such as an Overview, a set of requirements or a PRD.
- Applications – your code bases. Upload the source of a product and Next builds a map of it that agents and people can explore.
- Knowledge Bases – collections of documents, such as policies, manuals or research, that agents can read and quote from.
Together these are the context your agents work from. An agent pack decides who does the work. Applications and knowledge bases decide what they know while they do it.
How this fits the wider flow
Work in Paradym Next moves from an idea, through the Inbox and Portfolio planning, into the Intake Queue, then into Projects and work streams, and finally into Support. Context & Agents sits alongside that flow rather than inside it. It feeds the steps where Next writes on your behalf:
| Where it is used | What happens |
|---|---|
| Planning drafts | When you generate an item’s Overview, Requirements or PRD in the Portfolio, an agent pack writes the draft. Its agents read the knowledge bases they have been given. |
| Next Studio | Each Next Studio step (planning, code review, test review and code commit) can be run by an agent pack, set in Governance → Delivery → Next Studio. |
| Asking Next | When you ask Next a question from one of these pages, it knows which pack, application or knowledge base you are looking at. |
A few terms you will meet
You do not need a technical background to use these pages, but a handful of words appear often.
| Term | What it means here |
|---|---|
| Agent | An AI assistant with one job, such as writing requirements or checking a draft. |
| Agent pack | A set of agents that work together, in order, to produce one result. |
| Version | A saved edition of something. Packs, applications and documents all keep versions, so a change never silently overwrites what came before. |
| Code graph | The map Next builds from an application’s source: its files, the definitions inside them and how they connect. |
| Node | One item on a graph. In a pack it is a step run by an agent; in a code graph it is a definition, such as a function or a class. |
| Edge | A connection between two nodes, such as one function calling another. |
| Token | The unit AI models count text in, roughly three-quarters of a word. Token counts show how much work, and so cost, something took. |
| Chunk | A short piece of a document. Documents are cut into chunks so a search can find the exact passage it needs. |
| Embedding | A way of storing a chunk’s meaning so it can be found by what it says, not just the words it uses. |
| Enrichment | An extra pass after a document is added, in which Next writes summaries that help agents decide what to read. |
The Glossary has the full list.
Pages in this section
| Page | What it covers |
|---|---|
| Agent Packs | The list of agent packs in your workspace, and how to find one. |
| Agent pack detail | How a pack runs, step by step, with the agents, tools and knowledge bases each step uses. |
| Applications | The list of code bases Next can read, their status and versions, and how to add or delete one. |
| Application detail | One application: its versions and repositories, file summaries and linked knowledge bases. |
| Code files | Read an application’s source, search it and see what each file is for. |
| Code map | A picture of how the parts of an application depend on each other. |
| Code relations | Start from one definition and walk to everything it touches. |
| Knowledge Bases | The list of document collections, how to create, group and delete them. |
| Knowledge base detail | Upload documents, check their progress, and share a knowledge base with other workspaces. |
| Knowledge base document | One document as the agents read it, section by section. |
Typical workflow
- Create a knowledge base for a body of material, such as your product policies, and upload its documents. See Knowledge Bases.
- Check the documents finished processing on the knowledge base’s page. A document is searchable once its Ingest status reaches completed. See Knowledge base detail.
- Create an application for each code base Next should understand, and upload its source as a zip. See Applications.
- Link related knowledge bases to the application, so everyone can see which material belongs with it. See Application detail.
- Explore the code with the Files, Map and Relations tabs to check Next has read it the way you expect. See Code files.
- Review the agent packs your workspace uses, and which knowledge bases each step can read. See Agent pack detail.
- Choose which packs draft your plans in Governance → Planning → Agent Packs. See Agent Packs governance.