Glossary

Agent — a saved unit of work with a trigger, a workflow and a delivery destination. Agents overview

Agent spec — the trigger / workflow / delivery panel that fills in as you build an agent. Build an agent

Pending validation — the status of an agent that arrived from a colleague or a template, until the recipient completes a successful test run. Agents you build yourself are Active on save. Validation and review

Bot — an AI buyer in roleplay. Built from a LinkedIn profile, a CRM contact, or generated as synthetic. Build a bot

Certification — the final stage of a training plan, and the thing you can report on. Stages and pass gates

Credit — the unit of model consumption. Credits and usage

Data source — one connected origin of calls, CRM records or documents. What agents can see

Delivery — where an agent's output goes: Slack, email, CRM write-back, or a Live App. Delivery

Dimension — one axis of a scorecard, e.g. objection handling. Scoring overview

Document store — a named collection of documents that products read from. Document stores

Evaluation template — the object behind a scorecard, as labelled in the UI. Scorecard builder

Event trigger — a trigger that fires on something happening upstream, rather than on a clock. Triggers

Job — the scheduled unit behind a run: the thing that was due to fire. Runs and debugging

Live App — an HTML page an agent publishes at a stable URL, with data that stays live. Apps overview

MCP server — a Model Context Protocol server. Taizen's own lets clients reach your workspace (MCP server); external ones give your agents extra tools (MCP connections).

Meeting — a roleplay set up against one specific upcoming conversation, with a stakeholder roster. Meetings and prep

Pass gate — the score required on a stage before the next unlocks. Stages and pass gates

Run — one execution of an agent, with its inputs, steps, output, duration and cost. Runs and debugging

Score — the result of applying a scorecard to one call. Not the scorecard itself. Scoring overview

Scorecard — the rubric: dimensions, standards and weights. Scorecard builder

Session — one recorded roleplay, scoreable, replayable and shareable. Run a session

Skill — a reusable capability document telling agents how to do a recurring task, scoped Personal, Project or Global. Skills declare required inputs and can call other skills. Skills and tools

Stage — one step of difficulty in a training plan; contains steps. Stages and pass gates

Superseded — an earlier interview kept for the record after a re-trigger. Read the results

Test run — a builder-only run that produces output without delivering it. Build an agent

Timeline — every run, newest first. Runs and debugging

Training plan — a sequence of escalating roleplay stages with pass gates and a certification. Training overview

Trigger — what causes an agent to run: a schedule or an event. Triggers

Workflow — an agent's name, model and instructions. Build an agent