- A2A
- Agent2Agent: a protocol for agents built by different teams to discover each other and delegate work. It addresses a different layer than MCP rather than competing with it.
- Agent
- A system that decides its own next step and takes actions in other systems, rather than only producing text for a person to act on.
- Context window
- The maximum amount of text a model can consider at once, counted in tokens. Anything beyond it has to be selected, summarised or dropped.
- Discovery
- The conversations that precede a proposal, where you establish the client's problem, process, available data and constraints. While discovery is running, no architecture, timeline or price has been promised yet.
- Embeddings
- Numeric representations of text that place similar meanings close together, which is what lets a system retrieve passages by meaning rather than by keyword.
- Guardrails
- The checks placed around a model — on what goes in, what comes out and what it is allowed to do — that keep a working demo from becoming an incident in production.
- Inference
- Running a trained model to produce an answer. It is the recurring cost of an AI system, as opposed to the one-off cost of building it.
- LLM
- A large language model: a system trained to continue text. It manipulates information you give it; it does not hold your company knowledge unless you supply it.
- MCP
- Model Context Protocol: a common way to expose tools and data to a model, so an integration written once can serve any compatible client.
- Prompt injection
- Text a model reads as data and then obeys as an instruction. It arrives either in the chat message itself or from inside a document, email or web page the model was asked to read.
- RAG
- Retrieval-augmented generation: find the relevant passages first, then let the model answer using them. It supplies context, it does not train the model.
- System prompt
- The standing instruction an application places in front of a model: who it is, what it may do, how it should sound. It shapes answers strongly, but it is not a lock — the model reads it as text, alongside everything else it is given.
- Text2SQL
- Turning a question in plain language into a database query. It answers questions about numbers in a database, where RAG answers questions about text in documents.
- Token
- The unit a model reads and writes in — roughly a word fragment. Vendors bill per token, so tokens are the unit your invoice is denominated in.
- Tool calling
- Letting a model invoke a defined function — search a catalogue, create a ticket — instead of only describing what should happen.
- Vector database
- A store built to search embeddings quickly. It is where a retrieval system keeps the indexed fragments of your documents.
- Workflow
- A process with fixed steps decided in advance. When the steps never vary, a workflow is cheaper and more predictable than an agent.