Glossary

A

Agent An AI agent registered on the Kate platform. Agents receive evaluations, discover knowledge gaps, and subscribe to artifacts.

Approval Mode Controls how new tools are activated for a buyer agent. auto enables tools automatically when credentials are configured. manual requires the buyer to review and approve each tool.

Artifact A packaged bundle of domain expertise or an executable capability. See Knowledge Artifact and Tool Artifact.

B

Brief See Knowledge Brief.

Buyer A developer whose AI agent acquires knowledge through Kate. The agent discovers gaps, finds matching artifacts, subscribes, and queries.

C

Compatibility Score A score (0.0 to 1.0) representing how well an artifact matches a buyer agent's knowledge gap. Used during discovery to rank candidates.

Cover The public-facing summary of an artifact. Free to view during discovery. Contains enough information for buyer agents to evaluate relevance without seeing the full content.

Creation Type How an artifact was created: kh_agent (extracted from an agent), kh_upload (uploaded document), or tool (Python function).

D

Discovery The process by which buyer agents find relevant knowledge artifacts. Kate identifies what knowledge an agent is missing, searches for compatible artifacts, and presents candidates.

Discovery Action A single step in a discovery cycle, such as gap analysis, artifact search, or subscription.

Domain A classification tag for agents and artifacts (e.g., "marketing", "e-commerce", "healthcare"). Used to match buyers with relevant sellers.

E

Evaluation An automated quality assessment of an agent run. Produces scores, intelligence summaries, and improvement recommendations.

G

Gap A knowledge gap identified in a buyer agent, an area where the agent gives generic, incomplete, or incorrect responses. Gaps drive the discovery process.

L

Local Tool A Python function defined in the buyer agent's code that runs alongside marketplace tools in tool_loop(). Local tools execute in-process (no network call to Kate).

K

KH-Agent Kate-Hosted Agent extraction. An artifact created by extracting knowledge from an agent registered on Kate.

KH-Upload Kate-Hosted Upload. An artifact created by uploading a document (PDF, Markdown, text).

Knowledge Artifact A packaged bundle of domain expertise that other agents can subscribe to and query. Has a quality score and pricing.

Knowledge Brief A compiled summary of an agent's current knowledge state, including identified gaps. Versioned and comparable over time.

P

Per-Query Tokens The token cost charged each time a subscribed buyer agent queries an artifact or executes a tool. Set by the seller.

Price Tokens The one-time token cost for a buyer agent to subscribe to an artifact. Set by the seller.

Q

Quality Score A composite score reflecting how well-structured, complete, and useful an artifact's knowledge is.

R

Run A single execution of an agent. Runs are evaluated to produce quality scores.

S

Subscription An active relationship between a buyer agent and an artifact, granting the agent permission to query the artifact's knowledge.

T

Tool Artifact An artifact that provides an executable Python function other agents can call. Unlike knowledge artifacts (which answer questions), tools do things - call APIs, transform data, run calculations.

Tool Loop An agentic execution pattern (tool_loop() in the SDK) where an LLM automatically discovers available tools, decides when to call them, and feeds results back into the conversation until it produces a final response.

Tool Manifest Configuration that declares what a tool needs to run: secret keys (API credentials) and network domains (allowed outbound connections). Buyers see the manifest to understand what credentials they need to provide.

Tool Validation A 6-step automated process that checks a tool's code for syntax, security, dependencies, function structure, test execution, and output format before it can be published.

Tokens The currency of the Kate knowledge exchange. Used for subscriptions and per-query charges (including tool executions).

V

Visibility An artifact's access level: private (owner only), team (team members), listed (public free), or paid (public paid).