- Docs
- AI
- MCP & AI assistants
- Tools, prompts & commands
Tools, prompts & commands
You never call these tools yourself: you ask in plain words and the assistant picks them. Knowing them helps you ask well and understand what the assistant reports.
In short
On this page you can
- See which of the 18 tools only read and which can write.
- Start a ready-made prompt: Status, Modules, Scenarios, Red runs.
- Paste a command copied with With MCP or Copy MCP prompt.
- Understand the work keys the assistant reports, such as
SCENARIOSorHEAL.
How to use it
- Ask in plain words, naming the project.
- Let the assistant start with
smartest_contextorsmartest_status. - Follow the work key it suggests, riskiest target first.
- Approve or refuse each change it proposes.
The 18 tools#
Tools that only read run straight away. Tools that write follow the propose → apply rule or ask for a click, as noted.
Find out where you stand#
| Tool | What it does | Writes? |
|---|---|---|
smartest_context “Where are we” | The first call. Without a project, lists your projects so the assistant can ask which one. With a project: where it stands, what a module, requirement or scenario holds, what comes next, and the rules for that level. With a work key: its targets, riskiest first. | No |
smartest_status | The status of the project or a module: the Test Health numbers (coverage, funnel), the sub-requirements automation cannot prove, and the open work as work keys with counts, in priority order. | No |
smartest_discover “Discover what is known” | What SmartestQA already knows when there is no code, browser or document to read: pages with their actions, the saved API library, documents, the browser and device commands a script step may use, and what uses a data key. | No |
smartest_actions “Action catalog” | The actions (controls) element steps use, with their locators and how many scenarios use each; by page or module. | No |
smartest_case_rules | The rules for writing scenario steps, by topic (web, mobile, desktop, API, database, mail, AI steps, templates…), with a rules version the writer checks. | No |
Change records#
| Tool | What it does | Writes? |
|---|---|---|
smartest_propose “Propose a change” | Turns a draft into a diff with its impact, hints and a preview id. Levels: environment, module, main requirement, sub-requirement, note, document, page (purpose, roles), scenario plan, data keys, actions, defect, load test, API inventory. Intents: create, enrich (only adds), update, reorganize, remove. | No: it writes nothing |
smartest_apply | Writes exactly what the preview showed, after you approved it. | Yes, after your yes; removals need an Admin and a click |
smartest_action_write | Creates actions or changes an action’s locators, name or page. First call shows the diff and the scenarios each change touches; the second call, with only the preview, writes it. | Yes, after your yes |
smartest_api_library | Adds or changes a request in the API library. Tokens and passwords are data source references, never plain values. | Yes |
Write and run scenarios#
| Tool | What it does | Writes? |
|---|---|---|
smartest_write_case | Writes the steps of a planned scenario and dry-runs it on the server or on your online agent. All quality checks run before anything is written. If the run is red, it names the failed step and its cause; the third red run marks the scenario BLOCKED. | Yes (the scenario you approved in its plan) |
smartest_run_case | Dry-runs a written scenario again, steps unchanged, after you fixed a locator, data or the environment. | Starts a run |
smartest_run_status | The result of a dry run that was still running when the assistant stopped waiting. | No |
Pipelines#
| Tool | What it does | Writes? |
|---|---|---|
smartest_module_pipelines | Builds and fills a module’s standard pipelines (functional, API, security) and adds its scenarios to the project’s Smoke, E2E and Known defects pipelines, by their tags. | Yes |
smartest_pipeline | Lists pipelines, runs one, reads a run’s outcome (waiting up to 50 seconds), and sets a schedule for a pipeline that has passed once. It asks you the time and time zone. | Run and schedule: yes |
Results, maintenance and defects#
| Tool | What it does | Writes? |
|---|---|---|
smartest_runs “Runs, reports and logs” | Reports newest first, one run per scenario, a scenario run step by step with its failed step, locators and screenshot, pass/fail history, what runs now and runner health. It can stop a run, cancel a self-hosted runner job that has not started yet, and mail a report. It cannot stop a dry run: no one can. | Stop, cancel and mail ask for a click |
smartest_heal “Auto-heal decisions” | Healed locators waiting for a decision (the same list as in Maintain). Approve or reject one record: first a preview, then again after you agree. Reject restores the locators from before the heal. | Yes, after your yes |
smartest_defects “Tasks and defects” | Open defects, one in full, the company’s status names, comments and attachments; update, close (with what was done), comment, attach. A new defect is proposed with smartest_propose. Defects cannot be deleted, only closed. | Update, close, comment, attach |
smartest_integrations | Systems set up in SmartestQA: Jira, Xray and Confluence (find, map, sync, comment, export), the test mailbox (messages, wait, one-time codes, links), databases and Kafka (list, query, execute) and Excel sources. | Bulk sync and database execute ask for a click |
Ready-made prompts#
MCP clients that show prompts (Claude Desktop: the + menu of the chat) offer four. Each asks for the project id (and optionally a module id); leave it empty and the assistant asks which project.
| Prompt | Does |
|---|---|
| Status | Where the project or a module stands: the numbers and what comes first. |
| Modules | Groups the recorded pages into modules by business area. |
| Scenarios | Plans and writes scenarios for the sub-requirements that have none, riskiest first. |
| Red runs | Reads the red pipeline runs, classifies them, fixes them or proposes defects. |
Commands from the app#
Several buttons copy a ready request to your clipboard:
- With MCP on the Setup Roadmap: “Command copied. Paste it into Claude (or another assistant connected to the SmartestQA MCP).” MCP setup opens the MCP tab in Test Agents.
- Copy MCP prompt in Test Health: “Copies a ready prompt. Paste it into Claude or any MCP client and it does the work.”
The first line of such a request is a command the assistant recognises:
SmartestQA project 12 · SCENARIOS
SmartestQA project 12 · HEAL · module:4It means: in project 12, work on the key SCENARIOS (optionally for module 4). The assistant calls smartest_context with that key and starts with the riskiest target.
Work keys#
The assistant names open work with these keys, in smartest_status and in commands.
| Key | Means |
|---|---|
ENV | The environment is not complete (language, start URL, where runs happen) |
MODULES | Recorded pages that belong to no module |
REQUIREMENTS | Main requirements in no module, modules without a main requirement |
SUBS | Main requirements without a sub-requirement |
SCENARIOS | Sub-requirements without a scenario |
PROVE | Risky sub-requirements that are not proven |
ASSERT | Scenarios that check no result |
API | The module’s API inventory, endpoints without an API scenario, writes without a refusal scenario |
SECURITY | Critical endpoints without their security checks |
CLEANUP | Scenarios that create data and remove nothing |
HEAL | Red pipelines |
DEFECTS | Open defects (the assistant offers to close those whose scenario ran green) |
DATA | Data keys without a value, scenarios that refer to a missing key |
PIPELINES | Green scenarios in no pipeline, pipelines that never ran or have no schedule |
SMOKE | A module whose main path is not in Smoke |
GREEN | Ready scenarios that never passed in a pipeline run |
CURATE | Scenarios linked to no requirement |
PERF | Critical endpoints without a load test |
AGENTS | No autonomous agent watches the project |
STATUS | The status report |
How to read the assistant’s answers#
- Every tool answers ok, fail or ask. A text that starts with “X” is a failure: the assistant must say what failed and must not report it as done.
- ASK USER means the assistant has to ask you before going on, for example to approve a diff or to choose a project.
- NEXT: names the call that usually comes next. The assistant follows SmartestQA’s hints before its own ideas, and can go against a quality gate only by naming it and giving a reason.
- Tool results are in English. The assistant restates what matters in your language.
Didn’t find your answer?Write to support with your project name and what you see on screen.
info@smartestqa.ai