Welcome to bugAgent

We're thrilled to have you here! Dive in to learn how to give your AI agents their own bug reporting enrichment platform — along with a host of features and tools designed for the agentic era of software development.

bugAgent's mission is to 10x the velocity of testers by removing administration friction and letting them do what they do best — test in flow. Whether you're building autonomous coding assistants, integrating QA into CI/CD pipelines, or looking for a smarter way to manage bugs — you're in the right place.

Get Started

Jump into the area that fits your needs:

Why bugAgent? Our Philosophy

Context-driven testing. Measurable quality. Continuous improvement.

The Most Context-Aware Bug Reporting Platform

Every feature in bugAgent is designed to capture, preserve, and use context. Traditional bug trackers lose critical information at every handoff — between the tester who found the issue, the report that describes it, and the developer who fixes it. bugAgent eliminates those gaps by maintaining deep context across every touchpoint:

Quality of Testing = Quality of Product

bugAgent doesn't just track bugs — it actively measures and improves the quality of testing itself. The core thesis is simple: higher quality testing produces better bug reports, which leads to faster fixes and higher product quality.

The Continuous Improvement Loop

These capabilities create a virtuous cycle that raises product quality over time:

Measure testing quality → Identify gaps → Improve practices → Re-measure → Ship better software

Every report scored, every quality trend tracked, and every coverage gap identified feeds back into better testing practices. Teams using bugAgent don't just find more bugs — they get measurably better at finding them.

Introduction

Give AI agents a bug platform.

Software is increasingly written, tested, and maintained by AI agents. These agents need infrastructure — not just for writing code, but for the entire quality lifecycle. bugAgent is the platform that gives agents (and the humans who work alongside them) a shared language for bugs: one that's structured, enriched, and ready for action.

We built bugAgent because we believe the next generation of software teams will be hybrid — humans and agents collaborating in real-time. For that to work, bugs can't live in a spreadsheet or a GUI-first tracker. They need an API-first, protocol-native home where every participant (agent, human, CI pipeline, or IDE) can read, write, and act on them instantly.

What is bugAgent?

bugAgent is an API platform for giving AI agents their own bug reporting and product quality infrastructure. It goes beyond tracking — bugAgent is a report enrichment platform that classifies, contextualizes, and routes bugs, feature requests, enhancements, and more automatically.

With bugAgent, agents can:

• Manage reports autonomously — create, classify, query, update, and resolve bug reports, feature requests, enhancements, and more without human intervention. Each report includes a built-in timer you can start, stop, and resume to track exact testing time down to the second. Click the time display to manually adjust. The timer is available on both the creation form and the report detail page
• Enrich reports with AI — every report is auto-classified across 19 types — bugs (UI, performance, crash, security, logic, data, network, accessibility, compatibility), feature requests, enhancements, technical debt, documentation, DevOps, UX improvements, and integrations — with a confidence score
• Deploy human testers Coming Soon — request on-demand exploratory or Playwright automation testing from human QA teams
• Scan code and UI Coming Soon — run multi-agent swarms to check for abnormalities, vulnerabilities, and broken or outdated dependencies
• Sync to external tools — push enriched reports to Jira with mapped fields, priority, and labels. Playwright scripts auto-sync to GitHub repos via the GitHub integration
• AI Assistant — your QA command center — a context-aware assistant that goes far beyond report creation. Create and update bug reports, change status, severity, and type, add comments, create testing notes in 5 formats (Markdown, Plain Text, Checklist, Outline, Bug Template), list and search reports, notes, automations, and schedules, and send feedback — all through natural conversation. Use voice input powered by Whisper transcription, attach files, and let the AI analyze session replays to auto-draft reports. Start a new chat anytime with the New Chat button. The assistant runs in the authenticated dashboard; browser SDK captures hand off to it after login
• bugAgent Tools (SDK) — install the SDK with a project-bound, exact-origin ba_pub_ browser key created under Settings → Integrations. It captures the last 60 seconds of clicks, navigation, DOM mutations, errors, and network failures. Includes console log capture with full stack traces (last 50 entries), a network waterfall of failed and slow API requests, real-time performance metrics (page load, FCP, FPS, memory, long tasks), optional screen + voice recording via the FAB toolbar, DOM snapshot replay, and a browser markup/annotation tool for drawing circles, arrows, and freehand annotations on screenshots. Capture does not require a user login; replay review and report creation do. Available on the Enterprise plan
• Quality Score — measure your testing — every report is scored 1–10 to measure the quality of the testing, not just the severity of the bug. The score evaluates 10 context-driven dimensions: reproduction steps, expected vs actual, environment details, evidence, root cause analysis, impact assessment, context & history, heuristics & oracles, clarity & structure, and actionability. Teams track their average score trending over time in Analytics, and the Quality Testing Health Score (0–100) gives leadership a single metric for testing effectiveness. Visual badges (red 1–3, amber 4–6, green 7–10) help teams identify weak reports before development starts. Available via dashboard, API, and MCP
• Detect duplicates automatically — when a new report is created, text similarity analysis compares titles and descriptions against existing reports in the same project, flagging potential duplicates with match scores and direct links
• Stat-card definitions — Open Bugs counts reports created within the selected time range whose current status is still open (new, awaiting-triage, confirmed, in-progress, blocked, retesting, or reopened); it is not a point-in-time snapshot, which is why the count differs across the 7/14/30/90-day ranges. Trends compare against the preceding period of the same length. Hover any stat card for its definition.
• Analytics Dashboard — a full-screen suite with 12+ interactive charts: Bug Reports Over Time (stacked bars by severity), Quality Score Trend (line chart), Severity/Status/Type breakdowns (donuts + bars), Top Bug Reporters leaderboard, Automation Health (pass/fail), Time Tracking by day and category, Notes Created sparkline, and Resolution Time. Includes a Quality Testing Health Score gauge (0–100) with crown icon on the best-performing area. Filter by time range (7/14/30/90 days) and project. Available on the Enterprise plan. Also available via the REST API (GET /analytics) and MCP get_stats tool
• No-code web automation — record browser actions with zero code — AI generates production-ready Playwright scripts automatically. Self-healing locators: when a selector breaks mid-run, the runner asks Claude for a better one and retries the step automatically, so UI drift doesn't turn your suite red — and every heal is logged (assertions are never healed, so real regressions still fail). Run tests on virtual viewports (30+ emulated devices) or live real browsers via BrowserStack (Chrome, Firefox, Safari, Edge on Windows and macOS). End-to-end testing that builds, manages, and scales with your product. Toggle between Virtual and Live modes on each run or schedule. AI optimization powered by Sonnet 4. Scripts track up to 10 versions with one-click undo. Line-by-line pass/fail highlighting. The Enterprise plan includes an automation coverage mind map on the dashboard
• Notes & Project Wiki — capture testing observations in 5 formats (Markdown, Plain Text, Bug Template, Checklist, Outline), promote durable knowledge into project-scoped wiki pages, and switch between card, list, and table views for large knowledge bases. Includes voice-to-text dictation via OpenAI Whisper, a built-in timer you can start, stop, and resume to track exact testing time down to the second — click the time display to manually edit. Tracked time transfers automatically when converting a note to a bug report. Also includes file attachments up to 400 MB, private, project-team, or selected-teammate visibility, auto-save, version history, archive, and keyword search with author, date, tag, wiki, and folder filters. Available on all plans
• Time Tracking — log daily QA effort with categories (testing, bug-triage, automation, review, meeting). Create, update, and delete time entries per project. Filter by period (today, week, month), category, and team member. View daily, weekly, and monthly summaries in the analytics dashboard. Available via dashboard, API, and MCP on Enterprise
• Test Cases — a complete test management system inspired by TestRail. Create test cases with detailed steps (action + expected result) and drag-and-drop step reordering. Group them into test suites, create and assign test runs to team members with notifications, execute by marking each case pass/fail/blocked/skip with auto-save, instantly file bug reports from failures with full test context — bug reports link back to the test case and test run for full traceability. Re-run only failed cases, view reports with pass rate and date filtering, color-coded cards, and CSV export. Use Rapid Mode for fast test case creation. Available via dashboard, REST API, and MCP. Free usage is limited and supports URL references; AI generation, AI tag suggestions, Figma import, and file attachments require Enterprise
• Geo-Snap — enter any URL and capture screenshots from 50+ countries simultaneously. Verify localized content (currency, language, regional pricing), test geo-redirects, monitor competitors across markets, and check compliance with local regulations like cookie banners and privacy notices. Free plan: 1 country at a time with 10 saved screenshots. Enterprise: up to 5 countries simultaneously with unlimited saved screenshots. Available via dashboard and REST API
• Mobile Testing — upload APK (Android) or IPA (iOS) apps and run automated tests on real devices. Record secure login flows, import existing scripts, or upload a known-working Maestro YAML flow for native BrowserStack execution. Refine with AI hardens a recorded tour into a self-healing, device-certified script: an agent explores the app on a real device, layers in resilient locators (accessibility id, resource id, semantic labels), and certifies it with repeated green runs — and the script is yours to export and run anywhere. Every run produces video recordings, screenshots, and logs. Failures can create bug reports automatically or on demand from the run detail page. Schedule runs on multiple devices with Slack/email notifications. Enterprise only. Available via dashboard, REST API, and MCP tools
• Performance testing — Page quality audits (Performance, Accessibility, Best Practices, SEO) combined with load testing using virtual users (RPS, latency percentiles p50/p90/p95/p99). Mobile app performance testing on real Android/iOS devices via BrowserStack App Profiling — track CPU, memory, FPS, battery drain, and launch time with a Web/Mobile toggle. Core Web Vitals tracking, auto-bug creation when thresholds are breached (web scores or mobile metrics) with screenshot, recommendations, diagnostics, and quality scoring. CSV report export and archive/unarchive runs. Enterprise only
• Security scanning — OWASP ZAP + Nuclei web DAST, MobSF mobile binary analysis, and Semgrep code SAST. Findings categorized by severity (Critical, High, Medium, Low, Info) with CVSS scores, CWE IDs, and remediation guidance. Auto-creates security bug reports for Critical/High findings. Enterprise only
• Kanban Board — visualize bug reports across 9 status columns (New, Awaiting Triage, Confirmed, In Progress, Blocked, Resolved, Retesting, Closed, Reopened) with drag-and-drop status management. Dragging a card updates the status in real time and syncs bi-directionally to Jira instantly. Toggle between list and kanban views with a persistent preference. Each card displays severity, type, description preview, and timestamps
• Team Booster — scale your QA team instantly with pre-configured tester accounts. Fill in team size (1–10), location, duration, technical level, and budget — accounts are created in seconds and testers appear in Team Management with full platform access. Available on the Enterprise plan. You will not be charged until approval has been given. Also available programmatically via the REST API (POST /team-booster) and the MCP scale_team tool

This allows agents to assume their own identity and communicate via the universal protocol of bugs — with services, people, and other agents across any IDE or platform.

<script src="https://app.bugagent.com/sdk/bugagent-sdk.js" data-key="ba_pub_your_browser_key_here" data-position="bottom-right"></script>

What can autonomous agents do on this platform?

Agents can use the REST API and Model Context Protocol tools enabled for their workspace plan. They can file bugs, read reports, trigger allowed tests, analyze results, and update statuses. Workspace limits still apply, and a person remains responsible for target authorization, defect decisions, and release decisions.

The Problem

Traditional bug management systems were built for humans clicking through web interfaces. They assume someone will manually fill out a form, pick a category from a dropdown, assign a priority, and tag the right team. This model breaks down in the age of AI-assisted development:

The Agentic Development Cycle

Software development is shifting from a human-only process to a collaborative cycle between humans and agents. In this new paradigm, AI agents write code, review pull requests, run tests, and monitor production — all alongside human developers.

A QA AI Agent fits naturally into this cycle as the quality layer — continuously scanning for issues, filing enriched bug reports, and coordinating fixes. But for this to work, the QA agent needs a platform that speaks its language: structured data over APIs, machine-readable classifications, and protocol-native communication.

bugAgent is that platform. It sits at the intersection where agents, humans, and tools converge — providing a universal protocol for quality that works across any IDE, CI system, or communication platform.

The Solution

bugAgent is an API-first bug enrichment platform designed for agents. Every feature, from report creation to Jira sync, is available through structured APIs and the Model Context Protocol — not locked behind a web UI.

API-First Infrastructure

What does API-first mean for a bug management platform?

AI-Native Features

bugAgent doesn't just store bugs — it understands them:

Claude Integration

Self-healing bug analysis powered by Anthropic.

Connect your Anthropic API key to push bug reports to Claude for root cause analysis and fix suggestions. Claude analyzes each report and returns actionable insights including likely causes, affected areas, and recommended fixes. Available on the Enterprise plan.

How to Connect

Navigate to Settings → Integrations in the dashboard. In the Claude by Anthropic section, enter your Anthropic API key (starts with sk-ant-) and select your preferred model (Claude Sonnet 4, Claude Opus 4, or Claude Haiku 3.5). Click Connect to save.

Per-Project Configuration

Once connected, configure Claude on a per-project basis:

• Auto-push bugs to Claude — toggle this on to automatically send every new bug report to Claude for analysis as soon as it's created
• Custom instructions — add project-specific context such as your tech stack, architecture patterns, or areas of focus (up to 5,000 characters). Claude uses these instructions to provide more relevant analysis

Automatic & Manual Regenerate

Developer Notes fire automatically on every bug creation — no button to click. When the reporter later edits the description or adds/removes attachments, a green "Regenerate?" banner appears inside the Developer Notes card; clicking Regenerate now refreshes both the root-cause analysis and the Likely Fix Area against the updated content. A Regenerate button is also available in the header of the card at any time for a fresh evaluation.

API and MCP Access

Developer Notes is available through multiple interfaces:

• REST API — call POST /api/claude/push with a report_id to regenerate Developer Notes programmatically. The response includes the full analysis text and timestamp. See the API reference for details.
• MCP tool — use the push_to_claude tool from any MCP client (Claude Desktop, Cursor, VS Code, Claude Code). Pass a report_id and receive Developer Notes directly in your AI workflow. See the MCP documentation.
• Dashboard — the Regenerate button on any bug report detail page re-runs the analysis on demand.

Connecting to the MCP Server

The bugAgent MCP server is live at https://mcp.bugagent.com/mcp and can be used from seven different clients. For most clients you'll grab an API key from Settings → Developers → API Keys (keys start with ba_live_). For OAuth-aware hosts that require static client_id / client_secret upfront — the Claude.ai web app is the most common example — generate platform-agnostic OAuth credentials from Settings → Developers → MCP Connectors instead.

1. MCP Inspector (web UI) — best for first-time testing. Run npx @modelcontextprotocol/inspector on macOS or Windows. Set Transport to Streamable HTTP, URL to https://mcp.bugagent.com/mcp, Connection Type to Proxy. Open the Authentication tab and add a header with Name Authorization and Value Bearer ba_live_YOUR_KEY. Click Connect.
2. Claude Desktop (Mac + Windows) — edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows), add a bugagent entry under mcpServers with type: "http", the URL, and the Authorization header. Restart Claude Desktop.
3. Claude Code (CLI) — one command: claude mcp add --transport http bugagent https://mcp.bugagent.com/mcp --header "Authorization: Bearer ba_live_YOUR_KEY"
4. Cursor — Settings → MCP → Add new server → HTTP transport, URL https://mcp.bugagent.com/mcp, header Authorization: Bearer ba_live_YOUR_KEY.
5. VS Code + Continue extension — edit ~/.continue/config.json (Mac) or %USERPROFILE%\.continue\config.json (Windows), add an mcpServers entry with type streamable-http, URL, and headers.
6. Claude.ai (web) — or any OAuth-aware MCP host — in bugAgent: Settings → Developers → MCP Connectors → Generate connector (Confidential mode; paste the redirect URI your host requires). Copy the client_id + client_secret. Paste them into the host’s connector form along with authorization URL https://mcp.bugagent.com/authorize and token URL https://mcp.bugagent.com/token. For Claude.ai specifically: claude.ai/customize/connectors → Add MCP connector (its redirect URI is https://claude.ai/api/mcp/auth_callback). Sign in with Google or email/password for consent. This flow is used when the host requires static client_id / client_secret upfront instead of driving RFC 7591 dynamic registration.
7. curl / terminal — hit the endpoint directly with JSON-RPC 2.0 payloads. Required headers: Authorization: Bearer ba_live_..., Content-Type: application/json, and Accept: application/json, text/event-stream.

Full step-by-step instructions with copy-paste config blocks for every option are on the MCP page → Connect to the Server section.

The Self-Healing Cycle

bugAgent as the orchestration layer for autonomous software quality.

bugAgent connects QA, development, and AI into one continuous improvement cycle. When all components are active, software effectively heals itself — with humans reviewing and guiding every step.

1. Record — Capture Issues From Every Source

Issues enter the system through multiple channels, each producing structured, enriched bug reports:

• FAB tools (SDK) — the floating action button captures DOM recordings, console logs, network waterfalls, performance metrics, and screenshots directly from the browser. A logged-in workspace member reviews the session and AI draft before creating a report.
• Playwright automation — scheduled and on-demand test scripts detect regressions and failures automatically. Failed runs generate bug reports with video evidence and error details.
• AI assistant — the dashboard AI assistant lets authenticated team members create detailed bug reports through natural conversation, including voice input and file attachments. It also supports triggering web and mobile performance tests, listing runs, and reviewing results.
• MCP agents — AI agents running in Claude Desktop, Cursor, VS Code, or Claude Code file bug reports directly through the Model Context Protocol. CI/CD agents can report issues as part of the pipeline.
• Email — send bugs via email and they are automatically classified, filed, and synced to your project tracker.

2. Diagnose — Developer Notes

Every bug report is auto-analyzed the moment it is created. Developer Notes arrive without any manual trigger:

• Auto-generated on creation — every new bug report fires Developer Notes automatically. No per-project flag, no Send button. Uses the platform Anthropic key (no per-team Claude connection required).
• Manual regenerate — the Regenerate button refreshes the analysis. When the description or attachments change, a Regenerate? banner appears so Notes can be refreshed against the new content.
• Structured analysis — probable cause, affected areas, suggested fix with code-level specifics, verification steps, and risk assessment — plus a "Likely Fix Area" sub-block pointing at specific files.
• Programmatic access — push_to_claude via MCP or POST /api/claude/push via REST regenerates on demand. analyze_fix_area / POST /api/claude/fix-area does the same for the Likely Fix Area sub-block.

3. Automate — Verify Fixes Continuously

Playwright automation scripts provide the verification backbone:

• Scheduled runs — tests run on a cron schedule to catch regressions before users do
• On-demand runs — trigger tests via the dashboard, API, or MCP after applying a fix
• 30+ device targets — run tests on desktops, laptops, iPhones, Android phones, and tablets. Select the device when triggering manual or scheduled runs
• AI optimization — send any script to Sonnet 4 for a 12-point optimization that fixes selectors, wait strategies, assertions, error handling, auth patterns, mobile compatibility, and strict mode issues
• Version history & undo — every script tracks up to 10 previous versions. One-click undo reverts to the prior version. Versions are saved before manual edits, AI optimization, and script regeneration
• Line-by-line results — after a run completes, the script highlights passed lines in green and failed lines in red so you can see exactly which step broke
• CI/CD integration — run tests as part of your deployment pipeline and auto-file bugs on failures
• Video evidence — every run produces video results so teams can see exactly what happened

4. Heal — AI Agents Close the Loop

Connected AI systems can read Claude's analysis and act on it autonomously:

• Read analysis via MCP — agents call get_bug_report to retrieve the Developer Notes including root cause, suggested fix, and Likely Fix Area
• Apply fixes — coding agents (Claude Code, Cursor, custom agents) read the suggested fix and apply code changes directly
• Trigger re-verification — after applying a fix, agents call run_automation to trigger the relevant Playwright tests
• Update status — agents update the bug report status and add comments documenting what was changed and verified

5. Human in the Loop — Review, Approve, Guide

Humans remain essential at every stage of the cycle:

• Review — teams review Claude's analysis and AI-applied fixes before merging to production
• Approve — pull request reviews, test result sign-offs, and deployment approvals keep humans in control
• Guide — custom instructions, project configuration, and priority settings direct how the system behaves
• Override — any automated action can be paused, reversed, or manually corrected at any time

The Full Picture

When all components are connected, bugAgent becomes the orchestration layer for a full-circle autonomous QA pipeline:

Record → Diagnose → Automate → Heal → Record …

Each cycle produces better software. Bugs are caught earlier, fixes are more targeted, regressions are detected faster, and the AI learns from project-specific context. Teams spend less time on triage and more time building features — while bugAgent orchestrates the quality lifecycle around them.

bugAgent Skills

Integrations that connect external tools directly into bugAgent's context engine.

Skills are first-class integrations that plug external services into bugAgent. Each Skill enriches the context engine with data from tools your team already uses — syncing repositories, pulling analysis, bridging issue trackers, and more. Once enabled, Skills operate automatically in the background, keeping your QA data connected and up to date.

Available Skills

How to Enable Skills

Navigate to Settings → Integrations in the dashboard. Each Skill has its own connection flow — OAuth for GitHub and Jira, API key for Claude. Once connected, configure per-project settings such as auto-push, repository mapping, or default Jira project.

Building Custom Skills

Partners and developers can build their own Skills to extend bugAgent. Custom Skills connect external tools via OAuth or API keys, feeding data into the context engine alongside the built-in integrations.

• Submit an integration — reach out via the homepage or email support@bugagent.com to propose a new Skill
• Build programmatically — use the API reference to build Skills that interact with bugAgent endpoints for reports, automation, and project data

Free Migration

Switch to bugAgent with zero friction — we handle the heavy lifting.

Moving from another bug tracking or QA tool? The bugAgent team will migrate your data for free. We export your existing bug reports, test scripts, and project configuration, then import everything into your new bugAgent workspace.

What's Included

• Bug report export — we export your existing bug reports from your current tool and import them into bugAgent with full metadata preserved
• Playwright automation scripts — we walk you through creating Playwright automation scripts tailored to your product's test scenarios
• Integration configuration — we help you connect Jira, GitHub, Slack, and Claude so your workflow is fully operational from day one
• Team onboarding — we set up your team accounts, projects, and permissions to match your existing structure

Dedicated QA Team

Professional QA engineers embedded in your workflow.

For teams that want hands-on support, bugAgent offers a dedicated QA team service. Our professional QA engineers work directly inside your bugAgent workspace to review, enrich, and act on your bug reports.

What the QA Team Does

• Bug report review and enrichment — our QA engineers review incoming reports, add reproduction steps, verify severity, and ensure every bug meets a high quality bar before it reaches your developers
• Automation script development — we develop Playwright automation scripts for your product, covering critical user flows, regression scenarios, and edge cases
• Custom features and integrations — need a custom Skill, a bespoke reporting dashboard, or a CI/CD pipeline integration? Our team builds it for you
• Ongoing QA operations — continuous testing, triage, and quality monitoring tailored to your release cadence and priorities

Test Cases

bugAgent includes a complete test case management system inspired by TestRail, built for modern QA teams. Manage your entire testing lifecycle from case creation through hands-free execution to analytics and stakeholder reports.

Core Concepts

• Test Cases — individual test scenarios with a name, description, preconditions, and ordered steps. Each step has an action and an expected result. Drag-and-drop step reordering. Cases include priority (critical/high/medium/low), type (functional, regression, smoke, integration, performance, security, usability, exploratory), tags, estimated execution time, and URL references
• Folders — an organizational hierarchy for your case catalog (one folder per case). Up to 3 levels deep. Drag-and-drop reorder on the Cases-tab sidebar; order persists per team
• Test Suites — test plans that group cases (many-to-many). Up to 3 levels deep. Running a parent suite automatically includes every descendant sub-suite's cases, deduped
• Test Runs — an execution session snapshotted from a suite. Each result row records which sub-suite the case came from so results can be grouped on the run page
• Enterprise features — AI test-case generation, AI tag suggestions, Figma import, and test-case file attachments are available on Enterprise, not Free
• Reports tab — a windowed analytics view: KPI strip with prior-period deltas, pass-rate trend chart, failure analysis (failing / flaky / regressed / failing suites), suite health, coverage (staleness distribution), tester productivity, and one-click PDF export for stakeholder distribution

Hands-Free Run Execution

The run review page is a carousel with one case visible at a time. A jumpbar across the top shows one dot per case, coloured by status; click any dot to jump. Auto-advances to the next untested case on Pass / Block / Skip; Fail stays put so you can dictate an Actual Result and spawn a bug.

• Keyboard shortcuts — P Pass · F Fail · B Block · S Skip · ← / → navigate
• Voice control (Chrome / Edge / Safari) — click the mic in the carousel topbar, then say: "Pass", "Fail", "Block", "Skip", "Next", "Previous", "Add notes" (transcribes into the notes field), "Save notes", or "Voice off". A ? button next to the mic reveals the full command list
• Auto-completion — when the last untested case gets a result, the run auto-completes and the Export Results block (CSV / JSON) appears in the summary panel

Workflow

1. Organize cases into folders; create suites as test plans that pull from them
2. Create a test run from a suite (sub-suites auto-included); assign to a team member with a notification
3. Review the run via the carousel — keyboard or voice; results auto-save
4. Fail a case → dictate the Actual Result → one click creates a bug report with full test context linked back to the case and run
5. Re-run only failed cases after fixes are deployed
6. Open the Reports tab to see KPIs, failure analysis, suite health, coverage, and tester productivity
7. One-click PDF export of the Reports tab for stakeholders; CSV / JSON per run for audit trails

Ask the AI Assistant

The in-app AI assistant understands the entire test-case surface — folders, sub-suite expansion, the carousel, voice commands, the Reports-tab analytics, and data export. A few examples:

• "Create a test case for the user registration flow"
• "Create a Smoke/Auth folder and move my login cases into it"
• "Walk me through running my Billing suite hands-free"
• "How's our pass rate trending this month?"
• "What cases are flaky this month?"
• "Which suites have the worst pass rate?"
• "What haven't we tested in the last 60 days?"
• "Who on the team is overloaded with assigned runs?"
• "Export the results of run X as CSV"
• "Generate a QA stakeholder PDF for the last 30 days"

The assistant scopes every query to your team and never pulls data from teams you aren't a member of.

Full Traceability

Every bug report traces back to its origin: manual testing, automation failure, test case run, AI assistant, or an authenticated report created from a FAB SDK capture. The entire QA lifecycle is connected:

• Test case runs link to bug reports — when a case fails and you create a bug report, the report includes full context: test case name, steps, expected vs actual, and the test run it came from
• Bug reports link to test cases and automation scripts — every bug report shows its origin, whether it was filed from a test run, an automation failure, the AI assistant, or a reviewed SDK capture
• Automation failures create linked bug reports — when a Playwright script fails, the generated bug report links back to the automation script with line-by-line results
• End-to-end chain — test case → test run → bug report → fix → re-test. No broken chains, no lost context

Access

Test Cases are available through the dashboard, the REST API (full CRUD for cases, suites, folders, runs, and the Reports-tab analytics), and the MCP server (tools for AI agent integration including folders, sub-suite aware runs, and report analytics). Duplicate cases with one click, reorder cases within suites or folders via drag-and-drop, and filter by priority, type, status, tags, folder, or suite.

External Agent Execution

Hermes Agent and other execution workers can run one human-curated, project-bound suite without becoming the system of record. bugAgent creates an immutable plan snapshot, accepts retry-safe result batches, calculates the canonical summary, and keeps release and defect decisions with a person. Retrying with the same external_run_id resumes the existing run and does not consume another run. Deleting data does not reset monthly run usage.

The customer supplies and pays for the target browser, model, and network. Target credentials remain in the customer's execution runtime, and customers must restrict target access and network egress.

Start with the Hermes first-run guide, then use the public starter kit and community skill. This is a bugAgent-maintained community integration, not an official Nous Research integration.

Geo-Snap

Geo-Snap lets you enter any URL and capture screenshots of how a website appears from 50+ countries simultaneously. Powered by real geo-IP routing, every screenshot shows exactly what a visitor in that country would see — including localized content, regional pricing, language variants, and geo-specific redirects.

How It Works

1. Enter a URL — paste any public website address into the Geo-Snap input
2. Select countries — choose from 50+ countries (United States, United Kingdom, Germany, Japan, Brazil, Australia, and many more)
3. Capture — screenshots are taken from real IP addresses in each selected country and delivered within seconds

Key Use Cases

• Localization verification — confirm that currency, language, date formats, and regional content display correctly for each market
• Geo-redirect testing — verify that visitors are redirected to the correct regional version of your site (e.g., .co.uk, .de, .jp)
• Compliance checks — ensure cookie consent banners, GDPR notices, and other regulatory requirements appear in the correct regions
• Competitive monitoring — see how competitor websites look and price products across different markets

Plan Limits

Access

Geo-Snap is available through the dashboard and the REST API (list, create, delete).

Mobile Testing

Mobile Testing lets you upload APK (Android) or IPA (iOS) apps and run automated tests on real BrowserStack devices — Pixel, Galaxy, iPhone, iPad, and more. Record test actions in-browser, import an existing script, or upload a known-working Maestro YAML flow for native execution. Runs return video, screenshots, step results, and executor-specific logs. Failed runs can create comprehensive bug reports with failure snapshots and step breakdowns.

Record & Generate Scripts

Walk through your app on a virtual device directly in the browser. Log tap, swipe, input, and back actions as you go. When you stop recording, AI (Claude Sonnet 4) generates three production-grade scripts simultaneously:

During recording, bugAgent correlates each physical tap with a fresh Appetize UI-tree snapshot and prefers unique resource IDs or accessibility labels over coordinates. Appetize’s 0_resource_name_obfuscated placeholder is not a runnable locator: bugAgent omits it and requires a stable nearby ID or label. If the available evidence is stale, duplicated, conflicting, or coordinate-only, the event log marks the tap for review instead of guessing. Choose the intended selector or explicitly confirm the recorded point; unresolved taps block script generation and saving. Orientation changes are recorded as executable steps, and recordings are limited to 500 events.

Save any reviewed script as an automation and run it on real devices immediately.

Run an Existing Maestro Flow Natively

1. Upload the APK or IPA, then open its app detail page.
2. Click Upload Maestro and select one .yaml or .yml flow (maximum 512 KB). bugAgent checks the file immediately, lists YAML and flow errors with line and column numbers, and keeps Create Automation disabled until blocking errors are fixed. The flow must contain an appId header and a command document separated by ---. The appId must exactly match the selected app’s Android package or iOS bundle ID; example placeholders such as com.example.app and obfuscated Android resource IDs are rejected before a BrowserStack run starts.
3. Keep native Maestro commands and runtime values in the flow. Commands such as inputRandomText and copyTextFrom, and validated expressions such as ${maestro.copiedText} and ${output.value}, pass through unchanged.
4. For an authenticated flow, use a dedicated, least-privilege synthetic test account. Place only ${USERNAME} and/or ${PASSWORD} as the complete scalar value of an inputText command, then select a same-project login profile as the default. A password placeholder must immediately follow a tapOn target that identifies a password, passcode, passphrase, or PIN field; reveal, reset, and toggle targets are rejected. The stored YAML contains placeholders, never the profile values.
5. For reusable synthetic form data, create a same-project variable profile and reference uppercase keys such as ${DATA_EMAIL} or ${DATA_REGION}. Every referenced key must exist before dispatch. Profiles allow 1–100 strings, at most 4096 UTF-8 bytes per value and 65536 bytes total. Reserved credential/runtime names are rejected.
6. Open the created automation, choose a compatible Android or iOS device and, when needed, override its defaults. For data profiles, an omitted run selection inherits the automation default, null uses no profile for that run, and a same-project UUID overrides it. Schedules inherit the automation’s current default. bugAgent sends the YAML directly to the native Maestro executor without an Appetize tour, action conversion, or AI refinement.
7. Review the flow result and evidence. Native Maestro runs copy video and screenshots into private bugAgent run storage. Exact known credential values are filtered, while variable-profile values receive best-effort exact-match filtering from persisted textual evidence. Transformed, partial, encoded, or app-derived data values may remain. Authorized private visual evidence remains available and may show values rendered by the tested app. If credential redaction context is unavailable or sanitization cannot be proven safe, detailed credentialed text is withheld while run status and available visual evidence remain.

When you record through Appetize, bugAgent discovers and stores the runtime package or bundle ID for generated Maestro YAML. If a direct-upload app does not have one stored yet, its first validated native flow establishes the ID; later flows must match it. Generated flows start the selected app, preserve stable resource IDs, and use visible labels or device-relative points only when a stable ID is unavailable. Native Maestro accepts the two login-profile variables above only in validated inputText commands, accepts bounded DATA_* references backed by a selected variable profile, and preserves validated built-in runtime expressions rooted at maestro, output, console, Math, or JSON. Every other unresolved environment variable, misplaced credential placeholder, and every password-like literal is rejected. The flow must be self-contained: inline runFlow commands work, while external runFlow, runScript, and retry file references require a multi-file suite and are rejected in v1. Use a flow and app state that already run safely in your external Maestro setup. Native Maestro and legacy imported Maestro are separate execution modes: imported Maestro continues through the existing appium_actions path unless browserstack_maestro is explicitly selected through the REST API or MCP tool.

Login Profiles & Credential Vault

Create project-scoped identities such as Admin, Contributor, or Read-only User from Settings → Login Profiles or directly from an authentication panel. The same profile can be selected for Mobile, Web Automation, and Exploratory AI manual or scheduled runs. Secret values are encrypted immediately, are never displayed again, and are never returned by the API or MCP server. Private profiles are visible, usable, and manageable only by their active creator. Shared profiles are available to active members who can access the same project; the creator or an active workspace owner/admin can rename, rotate, or delete a shared profile, while only its creator can change visibility. Existing product-specific credential endpoints remain compatible. Deleting a profile clears future-use defaults and makes affected runs fail visibly until another profile is selected.

Variable Profiles for Synthetic Test Data

Variable profiles keep one native Maestro YAML flow reusable across test personas, regions, and datasets. Use uppercase keys such as DATA_EMAIL, DATA_REGION, or DATA_PLAN, then reference them as ${DATA_EMAIL} in the YAML. Profiles are bound to one project. Authorized project members can read their values, while only the active creator or an active workspace owner/admin can change, delete, bind, or use them. Automation defaults apply to immediate runs and schedules; an immediate run can explicitly select another profile or select none.

Record Login with Secure Input

1. Select a login profile before starting the recording.
2. In the event toolbar, click Record Login. While login capture is active, typed and pasted values are not persisted in the event log, UI-tree snapshots, console capture, recorded actions, or generated script. Password fields remain masked by the app/device UI; avoid exposing sensitive values elsewhere on screen while recording.
3. Before entering each value, choose Mark Username or Mark Password, then interact with that field. The recording stores only a username or password placeholder and the field locator.
4. Click Done to return to normal recording. Stop and save the automation when the rest of the flow is complete.

A selected profile is saved as the automation’s default only when the recording contains at least one correctly tagged login field. Untagged secure fields remain empty on replay.

Select Credentials per Run

When you click Run Now, choose a login profile alongside the target device. The run-time selection overrides the automation’s default profile, so the same recording or native Maestro flow can validate different roles without duplication. Profiles and automations must belong to the same project. Private profiles require their active creator; shared profiles require active access to that project. Recorded/imported action automations inject values into credential-tagged fields; native Maestro flows substitute only the referenced ${USERNAME}/${PASSWORD} variables. Scheduled runs use the selected schedule profile or the automation default and re-check the schedule creator’s access at dispatch.

Import an Existing Script

From the Automations tab, click Import Script, choose an Android app, name the automation, and paste the source. The importer supports Appium Python, Appium JavaScript/WebdriverIO, Maestro YAML, and Playwright mobile-web scripts. It preserves supported resource IDs, accessibility IDs, XPath, UiSelector, text, role, and test-ID locators, then reports any selectors or commands it could not map. Obfuscated Android resource-ID placeholders are skipped with a warning instead of becoming executable steps. Imported steps run verbatim without the Refine agent loop. Script import is Android-only during the beta period.

Refine with AI Allowlisted Beta

Refine with AI takes an Android recording and iteratively hardens its locators and actions on a selected real device. Start it from an automation detail page, choose the certification device, and provide as much intent as the flow needs:

• Goal — the outcome the tour must achieve, such as reaching an order-confirmation screen
• Waypoints — ordered milestones the flow should pass through, one per line
• Step expectations — assertions tied to specific recorded steps, checked against screenshots captured during certification
• Strict mode — when enabled, certification fails if the goal, any waypoint, or a confident step expectation is not satisfied. With strict mode off, intent guides the agent and is reported, but a missed intent check does not by itself block certification

The builder shows the transcript, agent stages, and current tour steps. You can send guidance or corrections while it works; messages queue until the next safe point. Click Stop to interrupt an active job. Its draft and event history are retained and the device session is released. Starting Refine again for the same automation resumes that job. When certification succeeds, review the draft and click Approve & save to make it the automation’s active recorded action set.

Supported Platforms

• Android — upload .apk files. Works directly for both recording and running on real devices (Pixel, Galaxy, OnePlus, Xiaomi).
• iOS — upload .ipa files for running tests on real iPhones and iPads. For recording, also upload a simulator build (.app zipped) — build from Xcode targeting “iPhone Simulator” and zip the .app bundle.

Update App Version

Replace your app binary with a new version without creating a new entry. Click Update App on the app detail page and upload the new APK or IPA. The old file is automatically deleted from storage to save space, and all linked automations and scheduled runs will use the new version on the next execution — no reconfiguration needed. If linked automations use login profiles, the caller must be authorized for every profile or be an active workspace owner/admin before changing the binary, package identity, or simulator target. For iOS apps, replacing the IPA also removes the simulator build — you’ll need to re-upload the updated simulator .app to continue recording.

Run on Real Devices

Click Run Now on any automation to launch it on a real BrowserStack device. Select from popular Android and iOS devices — the dropdown automatically filters by your app’s platform. Recorded/imported automations use the Appium 2.x W3C WebDriver executor; native Maestro automations use BrowserStack’s Maestro build and session APIs.

Run Detail Page

• Real-time polling — status auto-updates every 10 seconds while the test is queued or running, with a spinner banner
• Video recording — watch the full test execution on the device
• Searchable logs — Appium logs, device logs, and network logs with text search and match highlighting
• Duration, started, completed timestamps
• Rerun — one-click rerun with the same device and automation
• Archive — archive completed runs to keep the list clean. View archived runs with the status filter

Bug Reports from Failures

When a mobile test fails or errors, bugAgent can create a detailed bug report automatically or when you click Create bug ticket on the run detail page. Reports include:

• Error message in a code block
• Test step breakdown with pass/fail icons and counts (e.g. “3/5 passed”)
• Device, platform, OS version in an environment table
• Failure snapshot as the first attachment (last screenshot = failure state)
• Video recording of the full run
• Full logs in a collapsible section (up to 5,000 characters)

The bug report links back to the run detail page and the automation. The run listing shows a red “Bug Report” link after a report is linked. Each user can turn Create bug reports for my failed mobile runs off under Account Settings → Mobile Automation. That choice applies only to runs started by that user, including their scheduled runs; results and configured notifications still complete normally. With automatic creation disabled, an authorized user can decide per failed run by selecting Create bug ticket. Repeated requests return the already-linked report instead of creating duplicates.

Mobile Analytics

The Analytics dashboard includes a Mobile Testing row showing: total runs, pass/fail rate donut, runs by device bar chart, and Android/iOS platform split. Visible when you have mobile run data.

Scheduling & Notifications

Schedule mobile test runs using cron expressions with timezone support. Configure Slack or email notifications on failure so your team is alerted immediately when regressions are detected.

Setup

bugAgent SaaS customers do not configure BrowserStack credentials. The managed runner owns the device-provider integration. Upload apps and create automations through the dashboard, REST API, or MCP tools. Contributors and self-hosted operators configure provider credentials only in the private runner environment described by the repository deployment guide; never place them in a browser, mobile script, or public API request.

Compliance Center

Enterprise plan only. Automate SOC2, ISO 27001, and GDPR compliance evidence collection, config drift monitoring, access reviews, and cross-service security event tracking.

Compliance Relevance by Feature

Getting Started

1. Connect services — Navigate to Compliance → Connect Services. Add your Cloudflare, GitHub, Sentry, Supabase, and/or Railway credentials.
2. Collect evidence — Click "Collect Evidence" from the Compliance Center. This pulls security settings, vulnerability alerts, error trends, deployment history, and more from each connected service.
3. Monitor config drift — Go to Config Audit to initialize baselines. bugAgent checks every 6 hours and alerts you when security settings change (e.g., SSL downgraded, HSTS disabled, WAF rules removed).
4. Run access reviews — Generate quarterly access reviews from the Access Review page. Audits team members, API key usage, MFA status, and recommends revoking inactive credentials.
5. Track security events — The Security Events page shows a unified timeline of WAF blocks (Cloudflare), errors (Sentry), and auth events. Cross-service correlation groups related events.
6. Schedule collection — Set up automated weekly or daily evidence collection. bugAgent collects and stores structured evidence for auditor review.

Evidence Collection Services

• Cloudflare: SSL mode, min TLS version, HSTS, WAF rules, rate limiting, security level, WAF events (24h)
• GitHub: Dependabot alerts (critical/high/moderate), branch protection rules, recent commits, security advisories
• Sentry: Unresolved errors, event volume, error trends
• Supabase: User counts, storage buckets (public/private), auth status
• Railway: Deployment history, service count, failed deploys

Penetration Testing

Active penetration testing is available from Security → New Scan → Pentest tab. Six modules run rate-limited active attacks against your target:

• SQL Injection: Tests forms and URL parameters with 12 payloads (error-based, blind, time-based, NoSQL)
• Auth Bypass / IDOR: Forced browsing to admin paths, sequential ID access, JWT none-algorithm test
• Privilege Escalation: Role parameter tampering, HTTP method override
• SSRF Detection: Tests URL inputs for internal network access (metadata endpoints, localhost)
• Endpoint Fuzzing: Probes 40+ sensitive paths (.git, .env, backups, GraphQL, actuator)
• ZAP Active Scan: Full OWASP ZAP active scan with authentication support

Enterprise Limits

MCP Tools

Compliance features are accessible via the MCP server for AI-assisted workflows:

• collect_compliance_evidence — Trigger evidence collection from connected services
• check_config_drift — Run config drift detection
• generate_access_review — Create quarterly access review report
• get_security_events — Query security event timeline

Singapore & SEA Compliance

bugAgent is extending its compliance program for customers in Singapore and the wider Southeast Asia region. This summarizes our regional posture; the full policy set is maintained in our internal compliance framework and is available to Enterprise customers under NDA for vendor due diligence.

Singapore PDPA

We are aligning to Singapore's Personal Data Protection Act (PDPA). The majority of its data-protection obligations — consent, purpose limitation, protection, retention, access & correction — are already met through our GDPR program. Two areas receive Singapore-specific handling:

• Data breach notification — a notifiable breach is reported to the Personal Data Protection Commission (PDPC) within 3 calendar days of assessment, and to affected individuals where significant harm is likely.
• Data Protection Officer — a Data Protection Officer is designated per PDPA s.11(3). The DPO business contact dpo@bugagent.com is published in the privacy policy per s.11(5).

Data Residency

By default, bugAgent data is hosted in the United States. The PDPA does not mandate local storage, and cross-border transfers are covered by contractual safeguards with all subprocessors. For customers with strict data-residency requirements, an in-region or on-premise deployment is available on Enterprise plans — see the Cloud or On-Premise option.

Cybersecurity & Certifications Roadmap

• ISO/IEC 27001 — Information Security Management System certification, in progress.
• CSA Cyber Trust — targeting the Cyber Security Agency of Singapore's Cyber Trust mark at the Promoter (Tier 3) tier; closely aligned with ISO 27001.
• CSA Cyber Essentials — on the roadmap as an entry-level stepping-stone mark.

Operational security policies — patch management, logging & monitoring, backup & recovery, and a vulnerability-assessment / penetration-testing (VAPT) schedule — underpin these certifications. See the Trust Center for current status.

MAS TRM Alignment (Financial Institutions)

bugAgent is not directly regulated by the Monetary Authority of Singapore, but for Singapore-regulated financial-institution (FI) customers our existing controls are mapped to the MAS Technology Risk Management Guidelines (chapters 3–18 + Annex A) as a vendor self-attestation that supports the FI's outsourcing-due-diligence obligations. The posture is suitable for non-material outsourcing out of the box. For material outsourcing, we will negotiate contractual safeguards (audit and inspection rights, prior subprocessor-change notice, termination-assistance plan) on request. Contact security@bugagent.com for the alignment document under NDA.

Healthcare Services Act 2020 (Sectoral Addendum)

bugAgent is not a licensed healthcare service under Singapore's Healthcare Services Act 2020 (HCSA) and is not subject to HCSA in its own right. We do not integrate with the National Electronic Health Record (NEHR), HealthHub, or any MOH-administered digital-health platform. For healthcare-sector customers in Singapore, our sectoral addendum documents this boundary and provides recommended customer safeguards — principally, use synthetic or de-identified test data in any artifact uploaded to bugAgent. No identifiable patient data is required for any product function. Inadvertent uploads can be deleted on request via dpo@bugagent.com with a 24-hour priority SLA.

Spam Control Act 2007

Transactional emails (password reset, team invite, report-assigned notification, billing receipt) and service notifications (security advisories, scheduled-maintenance notices) sit outside the Singapore Spam Control Act. Marketing email is sent only after explicit opt-in, with a working unsubscribe link in every message; we send no marketing SMS or telemarketing to Singapore phone numbers, so the PDPA Do Not Call provisions are not engaged today. Any future Singapore marketing program will go through a documented pre-launch checklist before the first message ships.

Vendor Security Questionnaire

For procurement teams, we maintain a standing CAIQ-style answer set covering 17 control domains including a Singapore-specific section (PDPA, Cybersecurity Act, Computer Misuse Act, CSA Cyber Trust, MAS TRM, data residency). We complete CAIQ v4, SIG-Lite, VSA, or custom procurement questionnaires from this answer set on request — typical turnaround 5 business days, faster for active enterprise procurement. Email security@bugagent.com.

Security & Two-Factor Authentication

bugAgent supports two-factor authentication (2FA) for all email/password accounts. Users can enable 2FA from Settings → Security to add an extra layer of protection to their account.

Supported Methods

How It Works

1. Enable: Go to Settings → Security → choose Authenticator App or SMS → scan QR code or enter phone number → verify with 6-digit code.
2. Login: Enter email + password → redirected to 2FA verification page → enter code from your app or SMS → access granted.
3. Google OAuth: 2FA is skipped entirely. Google manages security for OAuth users.
4. Disable: Settings → Security → Disable Two-Factor Authentication.

Verification Page

The 2FA verification page at login features a 6-digit code input with auto-advance between digits, paste support, and auto-submit when all 6 digits are entered. If both methods are enrolled, users can switch between “Use SMS instead” and “Use authenticator app instead.”

Middleware Protection

The dashboard middleware checks the Authenticator Assurance Level (AAL) on every request. If a user has 2FA enabled but hasn’t completed AAL2 verification, they are automatically redirected to the verification page. This prevents bypassing 2FA by navigating directly to dashboard URLs.

Enterprise Security

Enterprise plan includes SSO/SAML, audit logs, on-premise deployment option, and multiple workspace support. Contact sales for custom security requirements.

Analytics Dashboard

The Analytics Dashboard is a full-screen, interactive reporting suite available on the Enterprise plan. It provides real-time visibility into your team's testing health, bug trends, and QA performance across every project.

Quality Testing Health Score

At the top of the dashboard, a large gauge displays your overall Quality Testing Health Score from 0 to 100. This score is computed from four equally weighted components:

The component with the highest individual score displays a crown icon, highlighting your team's best-performing area at a glance.

Chart Sections

Below the health score, the dashboard includes 12+ interactive chart sections, each providing a different lens on your QA data:

Filtering & Controls

Every chart responds to two global filters at the top of the dashboard:

• Time Range — select 7, 14, 30, or 90 days. All charts, the health score, and the leaderboard update instantly
• Project Filter — scope the entire dashboard to a single project, or view aggregated data across all projects

API Access

All analytics data is available programmatically:

• REST API — GET /analytics returns daily counts, severity/status/type breakdowns, and health score. See the API Reference for API conventions
• MCP — the get_stats tool returns the same data for AI agents. Pass a days parameter (up to 90) to control the lookback window

Why bugAgent?

Here's why teams and developers choose bugAgent:

Developer Resources

Everything you need to build with bugAgent:

Frequently Asked Questions

Quick answers to the most common questions about bugAgent.

name: Test & Report
on: [push, pull_request]

jobs:
  test:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
        with:
          node-version: '22'

      - name: Install dependencies
        run: npm ci

      - name: Run tests
        id: tests
        run: npm test
        continue-on-error: true

      - name: Report failure to bugAgent
        if: steps.tests.outcome == 'failure'
        env:
          BUGAGENT_API_KEY: ${{ secrets.BUGAGENT_API_KEY }}
          REF: ${{ github.ref_name }}
          RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
        run: |
          curl -sS -X POST https://app.bugagent.com/api/reports \
            -H "Authorization: Bearer $BUGAGENT_API_KEY" \
            -H "Content-Type: application/json" \
            -d "$(jq -n \
              --arg title  "Test suite failed on $REF" \
              --arg desc   "CI run failed. See $RUN_URL" \
              '{title: $title, description: $desc, type: "functional", severity: "high", environment: "ci"}')"

      - name: Fail the job
        if: steps.tests.outcome == 'failure'
        run: exit 1

curl -sS -X POST https://app.bugagent.com/api/automations/run \
  -H "Authorization: Bearer $BUGAGENT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"automation_id": "your-automation-id"}'

CRITICAL=$(curl -sS \
  "https://app.bugagent.com/api/reports?severity=critical&status=open" \
  -H "Authorization: Bearer $BUGAGENT_API_KEY" \
  | jq '.reports | length')

if [ "$CRITICAL" -gt 0 ]; then
  echo "Blocked: $CRITICAL critical bug(s) open. Resolve before deploying."
  exit 1
fi

Support

We're here to help you build with bugAgent. Reach out anytime: