--- title: Agent gateway | Qventis One description: Coding agents create, update, select, run, triage and heal tests through MCP on the Qventis Engine, under your policy, with people approving every change. --- [Skip to content](https://qventis.ai/platform/agent-gateway#main) [![qventis](https://qventis.ai/hubfs/raw_assets/public/qventis-one/images/qventis-logo.svg)![qventis](https://qventis.ai/hubfs/raw_assets/public/qventis-one/images/qventis-logo-reverse.svg)](https://qventis.ai/?hsLang=en) - Platform [**Qventis Engine**The core of the platform. Automate any app, any stack, in plain English, with zero AI tokens at run time.WebDesktopTerminalAPIDataOne plain-English suite](https://qventis.ai/engine?hsLang=en) Platform [**Qventis One**The engine, scaled to one pane of glass](https://qventis.ai/one?hsLang=en) [**Release Intelligence**Every vendor change, mapped to your processes](https://qventis.ai/release-intelligence?hsLang=en) [**App Model**One model of your app, learned once](https://qventis.ai/one/app-model?hsLang=en) [**Quality View**Evidence and a signed go or no-go](https://qventis.ai/one/quality-view?hsLang=en) Automation [**Studio**Anyone can automate, in plain English](https://qventis.ai/platform/studio?hsLang=en) [**Eight engines**Functional to accessibility in one flow](https://qventis.ai/one#engines) [**Integrations**CI, ALM, chat and coding agents](https://qventis.ai/platform/integrations?hsLang=en) [**Trust and security**AI on your terms, people approve](https://qventis.ai/platform/trust?hsLang=en) How it works[Detect](https://qventis.ai/release-intelligence?hsLang=en)→[Assess](https://qventis.ai/one/app-model?hsLang=en)→[Automate](https://qventis.ai/engine?hsLang=en)→[Assure](https://qventis.ai/one/quality-view?hsLang=en) - [Release Intelligence](https://qventis.ai/release-intelligence?hsLang=en) - Platforms [**Release in confidence**Certify every vendor update and configuration change before your business feels it.PreviewTestProdNextImpact known before the window opens](https://qventis.ai/platforms?hsLang=en) Enterprise platforms [**Oracle**Quarterly updates](https://qventis.ai/platforms/oracle?hsLang=en)[**Workday**R1 and R2](https://qventis.ai/platforms/workday?hsLang=en) [**SAP**S/4HANA releases](https://qventis.ai/platforms/sap?hsLang=en)[**Salesforce**Three a year](https://qventis.ai/platforms/salesforce?hsLang=en) [**ServiceNow**Family releases](https://qventis.ai/platforms/servicenow?hsLang=en)[**Dynamics 365**Release waves](https://qventis.ai/platforms/dynamics-365?hsLang=en) [**nCino**Two calendars](https://qventis.ai/platforms/ncino?hsLang=en)[**Coupa**Major releases](https://qventis.ai/platforms/coupa?hsLang=en) [**All 40+ platforms**ERP, HCM, CRM, ITSM, banking, insurance](https://qventis.ai/engines/packaged-apps?hsLang=en) By need [**Release testing**Test only what a change affects](https://qventis.ai/solutions/release-testing?hsLang=en) [**Industries**Regulated and complex businesses](https://qventis.ai/industries?hsLang=en) [**Legacy desktop**Apps automation never reached](https://qventis.ai/solutions/legacy-desktop?hsLang=en) [**Quality CoE**One standard across the business](https://qventis.ai/solutions/quality-coe?hsLang=en) Certify an upgrade[Release notes in](https://qventis.ai/release-intelligence?hsLang=en)→[Impact traced](https://qventis.ai/release-intelligence?hsLang=en)→[Tests run](https://qventis.ai/engine?hsLang=en)→[Signed off](https://qventis.ai/one/quality-view?hsLang=en) - Resources [**Brochures and guides**Platform briefs sent to your inbox](https://qventis.ai/resources?hsLang=en) [**Customers**How teams release in confidence](https://qventis.ai/customers?hsLang=en) [**Blog**Field notes on releases, quality and AI](https://qventis.ai/blog?hsLang=en) [**Editions**Start with one need, grow into One](https://qventis.ai/editions?hsLang=en) [**Support**Guides and community for customers](https://qventis.ai/support?hsLang=en) - Company [**About qventis**Why we built one platform](https://qventis.ai/about?hsLang=en) [**Trust center**How we protect your data](https://qventis.ai/platform/trust?hsLang=en) [**Contact**Talk to a person](https://qventis.ai/contact?hsLang=en) Talk to usSend a request Search platforms, releasesCtrl K [Book a demo](https://qventis.ai/contact#demo) [Platform](https://qventis.ai/one?hsLang=en)/Agent gateway # Agentic coding, with tests that keep up Coding agents such as Claude Code, Copilot and Cursor use Qventis MCP tools across the whole test workflow: draft tests from a story or pull request, update them with the code, pick the smallest safe set, run them on the Qventis Engine, triage failures and propose heals. People approve every change. [Book a demo](https://qventis.ai/contact#demo) Get the brief - The full test workflow, not just runs - No agent approves itself - Every call audited **Agent gateway**MCP, CLI and API Coding agentIDE, PR or CIMCP clientQVENTIS MCP TOOLScreate\_test\_from\_storyDraft plain-English tests from a story or PRupdate\_tests\_for\_diffUpdate steps the code change affectsselect\_tests\_for\_changePick the smallest safe set for this changerun\_suiteRun on the engine, zero AI tokenstriage\_failuresGroup failures by cause with evidencepropose\_healSuggest a fix, a person approvesgenerate\_test\_dataReserve or synthesize masked recordsquery\_quality\_pulseRead risk, readiness and alertsApprovalA person signsevery changeEngineDeterministic runevidence keptQuality ViewPulse, alertsand audit trail Highlighted tools change tests and need approvalAgents propose, people approve ## What agents can do with your tests Every tool works on the same App Model and plain-English steps your testers read, so agent work stays reviewable. ### Create tests from stories and PRs Draft plain-English tests from acceptance criteria or a diff, ready for review. ### Update tests with the code When a change renames a field or moves a step, the agent proposes the matching test update. ### Select the smallest safe set Pick the tests a change touches, ranked by risk, instead of running everything. ### Run and read evidence Runs happen on the engine with zero AI tokens. Agents read logs, screenshots and expected versus actual results. ### Triage by cause Failures grouped into product defects, test issues and environment problems, with defects filed and evidence attached. ### Propose heals, never approve them Fixes arrive with before and after evidence and wait for a person. ### Skills you can give your coding agents 1. **01**Write a plain-English test 2. **02**Map a diff to impacted steps 3. **03**Generate masked test data 4. **04**Explain a failure 5. **05**Review a proposed heal 6. **06**Check release readiness ## Security built into every call The flowFrom change to approval IdentityScoped sign-in per agent PolicyChecked on every call AuditEvery call on record Three doorsCLI, HTTP API and MCP ### An agent verifies a change A coding agent finishes a change and checks whether it broke anything. Every step is checked against policy. - Signs in with its own scoped token - Finds and runs the tests its change affects - Reads the evidence and proposes a heal - A person approves or rejects the heal [Read: coding agents need a test gate](https://qventis.ai/blog/agents-need-a-test-gate?hsLang=en) ### Sign-in and allow-lists Agents sign in with SSO and OAuth through your identity provider. You decide which actions each may take, and where. - Scoped access for each agent - Allow-lists per project - Anything not on the list is denied [See Trust & security](https://qventis.ai/platform/trust?hsLang=en) ### Policy and approvals Each call is checked against role and attributes, the same way a person's request would be. - Deleting tests needs a person - Changing suites or shared components needs approval - Heals and releases stay with people [Read: earned autonomy for AI in testing](https://qventis.ai/blog/earned-autonomy?hsLang=en) ### Every call audited Who asked for what, with which inputs and what result, exported to your SIEM. - The same audit trail as people - Runs tied to the change behind them - Results in the same Quality View [See Quality View](https://qventis.ai/one/quality-view?hsLang=en) ### Three doors, one policy People, pipelines and agents reach the platform in different ways. The rules behind each door are the same. - CLI for engineers and CI jobs - HTTP API for pipelines, portals and scripts - MCP for coding and chat agents [See integrations](https://qventis.ai/platform/integrations?hsLang=en) ## Questions Which agents work with it? Any agent that speaks MCP, including Claude Code, GitHub Copilot and Cursor. CI pipelines use the CLI or HTTP API under the same policy. Does an agent spend AI tokens when it runs tests? Only its own, deciding what to call. The run itself uses zero AI tokens. Can we limit an agent to one project? Yes. Per-project allow-lists decide what an agent may do and where. ## Leave with one agent gated by policy See one agent connected to one project, running tests under your policy. [Book a demo](https://qventis.ai/contact#demo) Get the brief [![qventis](https://qventis.ai/hubfs/raw_assets/public/qventis-one/images/qventis-logo.svg)![qventis](https://qventis.ai/hubfs/raw_assets/public/qventis-one/images/qventis-logo-reverse.svg)](https://qventis.ai/?hsLang=en) [Support](https://qventis.ai/support?hsLang=en)[Trust](https://qventis.ai/platform/trust?hsLang=en)[Privacy and terms](https://qventis.ai/legal?hsLang=en)Contact © 2026 qventis.aiThird-party product names are trademarks of their respective owners.