Lists the full skill-tag catalogue filtered to tags compatible with a given scenario type (info_exchange or interview). Use recommend_scenario_tags or search_scenario_tags for a ranked shortlist when the user gave natural language or the list is too large. Returns each tag’s name, category, and description, and flags QnA tags that must be the only tag.

Suggests skill tags for a scenario type using the same Workbench pipeline as the Scenario Generator (embeddings plus LLM, up to about 20 tags). Use when the user describes a scenario or agent and wants a ranked shortlist before generate_scenario. Prefer this over list_compatible_tags when the catalogue is large or the user gave natural-language context. Read-only (no generation credits). Provide context_prompt and/or agent_uuid (at least one). On recommender timeout or unavailability, try search_scenario_tags or list_compatible_tags.

Finds skill tags for a scenario type by embedding similarity to a natural-language query (no LLM). Cheaper than recommend_scenario_tags — good for “find tags about X”. Read-only. A per-user tag-search rate limit applies in addition to the workspace public API limit; on rate limit, honor Retry-After, wait and retry, or use list_compatible_tags for the full filtered catalogue.

Registers an AI agent in VerifyAX given its name, connector type (A2A, API, DIRECTLINE for Copilot Studio, or MCP), URL, and optional auth or connector settings. For A2A agents it first verifies the agent card is reachable; DIRECTLINE and MCP agents are probed before creation. Returns the new agent’s uuid and whether connectivity was checked. Note: any token, Direct Line secret, or password passed here transits the conversation — prefer supplying credentials out of band where possible.

Lists the AI agents registered in your VerifyAX workspace, optionally filtered by connector type (A2A, API, DIRECTLINE, EXTENSION, or MCP). Returns each agent’s uuid, name, type, and URL.

Permanently deletes an agent from your VerifyAX workspace by its uuid. This cannot be undone. Returns confirmation of the deletion.

Generates a new test scenario of a given type (info_exchange or interview) with optional skill tags and context. Invalid tag names or incompatible tags fail immediately with 400/422 before a job is queued; use recommend_scenario_tags, search_scenario_tags, or list_compatible_tags to pick valid tag names. Typically takes 30s–2min; task-capable MCP clients receive a pollable task handle immediately, while others block until generation finishes. Set num_scenarios greater than 1 for batch mode (requires tag_pool). Returns the new scenario’s uuid, or batch uuids when batching, or a structured error if generation fails for other reasons after the job starts.

Lists the test scenarios in your VerifyAX workspace, optionally filtered by type (info_exchange or interview) and status. Returns each scenario’s uuid, name, type, and status.

Permanently deletes a scenario from your VerifyAX workspace by its uuid. This cannot be undone and fails if simulation runs still reference the scenario. Returns confirmation of the deletion.

Runs an agent against a scenario and evaluates the result end to end. Typically takes 30s–30min; task-capable MCP clients receive a pollable task handle immediately, while others block until the evaluation completes. Give it an agent uuid and a scenario uuid; it previews cost, runs the simulation, waits for it, and returns the evaluation scores. Optional timeout_minutes (1–240) overrides the scenario default for this run.

Lists recent simulation runs in your VerifyAX workspace, optionally filtered by status, agent, scenario, date range, search text, or run group. Returns each run’s uuid, status, agent, scenario, and evaluation handle.

Fetches the full details of a single simulation run by its uuid, including its status and the evaluation results when they are available. Use after a run to inspect scores and outcome.

Summarizes VerifyAX usage events over an optional time range or for a specific simulation, scenario, or job. Paginates across all matching events (up to a cap) and returns the total event count, a breakdown by product area, and total platform spend in USD when the API reports it.

Estimates the credit cost of running an agent against a scenario before triggering it. Returns the estimated credits, your current balance, and any pending committed spend. Optional timeout_minutes (1–240) affects the run-cost estimate.