Term lookup
Claude Ecosystem Glossary
A searchable glossary for Claude and Claude Code terminology, designed to send each query to the most relevant network site.
Claude Code
Anthropic coding-agent surface for reading repositories, editing files, running commands, and automating development workflows.
Open owner pageCLAUDE.md
A project instruction file used to give Claude durable context about commands, conventions, architecture, and workflow expectations.
Open owner pageSkill
A packaged set of instructions, scripts, examples, or assets that helps Claude handle a recurring task reliably.
Open owner pageMCP
Model Context Protocol, a standard pattern for connecting models to tools, data, and context providers.
Open owner pageHook
A Claude Code lifecycle automation point used to run checks, notifications, guards, or other actions around agent work.
Open owner pageSlash command
A reusable command-style prompt file for invoking a specific Claude Code workflow with predictable instructions.
Open owner pagePermissions profile
A settings pattern that allows or denies Claude Code tools based on the risk posture of a project or task.
Open owner pageContext window
The finite amount of text, tool output, files, and conversation history a model can consider at one time.
Open owner pageCompaction
A summarization or compression step that preserves important session state while reducing how much conversation history remains active.
Open owner pagePrompt caching
A cost and latency optimization for reused prompt prefixes, long instructions, examples, or stable documents.
Open owner pageMemory
Durable context that helps Claude preserve preferences, project rules, or prior work across interactions where supported.
Open owner pageArtifact
A substantial Claude output shown in a dedicated workspace, such as a document, code file, visualization, or mini-app.
Open owner pageProject
A Claude workspace with its own chats, uploaded knowledge, and instructions for a related body of work.
Open owner pageConnector
A permissioned integration that lets Claude retrieve data or act in another service under the connected account permissions.
Open owner pageSession log
A structured record of an agent session, including messages, tool calls, errors, and sometimes token usage.
Open owner pageJSONL transcript
A line-delimited JSON format often used for session traces where each line is one event or message record.
Open owner pageRelease radar
A recent-change view for Claude model, product, API, policy, or pricing updates over a chosen time window.
Open owner pageDeprecation
A model, API, or product surface moving toward retirement or replacement, usually requiring migration planning.
Open owner pageBenchmark
A standardized measurement task that can help compare models, but only if the metric, dataset, and saturation risk are understood.
Open owner pageRefusal
A model response that declines some or all of a request, usually because of safety, legal, privacy, or unsupported-claim boundaries.
Open owner pageHallucinated citation
A reference that looks scholarly or sourced but does not exist, does not support the claim, or has been misquoted.
Open owner pageExperiment loop
A repeatable cycle for an agent or researcher to propose, run, analyze, and decide on the next experiment.
Open owner pageSampling temperature
A generation setting that affects how concentrated or varied token choices are in a model output.
Open owner pageStyle guide
A reusable writing specification for voice, tone, perspective, formatting, banned words, and examples.
Open owner pageFAQ
What makes this different from the resources page?
The resources page organizes official sources; the glossary explains terms and routes each term to the best Claude Network site.
Does the glossary replace official Claude docs?
No. It is a navigation aid. For current product behavior, pricing, API details, and policy, verify the linked official source.
Can glossary search leave my browser?
No. The search box only filters the entries already loaded into the page.