SmartBear
Also known as: SmartBear Software, SmartBear BearQ, BearQ, SmartBear MCP Server, smartbear.com
SmartBear makes testing, API and observability software, and its BearQ product runs QA agents that test an application on their own. An Explorer agent maps the app's pages and flows, a QA Lead agent writes and edits tests and reports, and a Tester agent runs journeys and checks results, on a schedule, from a CI pipeline or from a coding agent. Teams choose how far the agents act alone and approve new regression tests before they join the suite.
SmartBear, founded by Jason Cohen in 2003 and based in Somerville, Massachusetts with a European headquarters in Galway, sells software quality tools to more than 32,000 organizations. Its portfolio spans Swagger and ReadyAPI for API design and testing, TestComplete, Reflect, Zephyr and QMetry for test automation and management, the Pact contract testing project, and BugSnag and AlertSite for error and synthetic monitoring. Customer logos include Atlassian, BMW, JetBlue, Microsoft, Meta, ZoomInfo and Cisco, and SmartBear counts 75% of the largest financial companies among its users.
BearQ is SmartBear's agentic QA product. Three agents share the work: the Explorer navigates the application and builds the Application Model, a living map of its pages, elements and workflows; the QA Lead creates and edits tests, answers questions and writes reports; and the Tester runs end to end user journeys across the UI and APIs and checks results against expected outcomes. When the app changes, the agents explore the changed areas again and revise tests.
Teams add context, such as how users sign in, documentation links and areas to avoid, and each task shows when that context was applied. Workflows chain native tasks with integration steps such as opening a Jira issue or syncing to Zephyr, branch on test results and run on a schedule, from a CI pipeline through the public API, from a coding agent through the MCP server or on demand. At Follett Software, a senior test engineer expects BearQ to save 75% of her time, reviewing tests written overnight and approving them for regression.
Autonomy is a setting. In human directed mode a team triggers each action and approves proposals before they are final, and proposed tests never become regression tests without approval. In autonomous mode BearQ explores, generates tests and validates on its own within scope limits, exclusion rules and time bounds, and both modes keep an audit trail of every action. The SmartBear MCP Server brings the same tools into AI assistants and IDEs, so a coding agent can trigger and monitor test runs, pull BugSnag error context and fetch OpenAPI specifications. Swagger can also generate MCP servers from an API.
BearQ is priced on agent compute units, with more intensive work using more units, unused units rolling over for the contract term and a free seven day trial with every capability open. SmartBear holds ISO/IEC 27001 and SOC 2 reports for products such as Swagger, Zephyr, PactFlow and QMetry, and BearQ offers role based access controls. BearQ suits product and QA teams that release often and want coverage beyond their scripted suites, especially those already on Jira and Zephyr. Teams that need to pick the models behind the agents or keep testing inside their own network will find less here.
Vendor details
Canonical URL
https://smartbear.com/
Category
Agent infrastructure
Subcategory
Autonomous QA agents and MCP access to testing and API tools
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
BearQ connects to Jira, Linear, Slack, GitHub, CI/CD pipelines, Zephyr and other SmartBear tools, with a public REST API and an MCP server that also reaches Swagger, BugSnag, QMetry, TestComplete and Reflect.
In practice
After a release, the Explorer remaps the changed checkout flow and the Tester finds a wishlist quantity bug no existing test covered.
A nightly workflow runs regression journeys, opens Jira issues for failures and syncs results to Zephyr.
A developer's coding agent calls the MCP server to run the affected tests and pull BugSnag errors before merging a fix.
Sources & related URLs
Related / legacy domains
Research sources
Agentic Index coverage score
10.5 / 14 capabilities · 75%
| Integrations & Tool Calling | Full |
|---|---|
|
BearQ workflows mix native tasks with steps that act in connected tools, such as opening a Jira issue or syncing results to Zephyr, and external agents reach issue trackers, collaboration tools and other SmartBear products. Connections cover Jira, Linear, Slack, GitHub and CI/CD pipelines. Through the MCP server, an assistant can trigger SmartBear test runs, pull BugSnag errors and fetch Swagger specifications. Sourcesupport.smartbear.com/bearq/docs/en/workflows.htmlread 2026-11-20 |
|
| Workflow Orchestration | Full |
|
Three agents divide the work: the Explorer maps the application, the QA Lead writes, edits and reports on tests, and the Tester runs journeys and checks outcomes. A workflow is a graph of steps that can include tasks for any of them alongside integration steps, branch on test results and log each run step by step. Saved workflows start and cancel through the API as well as in the product. Sourcesupport.smartbear.com/bearq/docs/en/bearq-agents.htmlread 2026-11-20 |
|
| Knowledge Grounding & RAG | Full |
|
The Explorer keeps the Application Model, a living map of the customer's pages, elements and workflows that the other agents use to choose and author tests, and it updates as the application changes. Teams add context on top: one global context for every task plus targeted knowledge items, such as documentation links and sign in rules, that apply only when relevant, with each task showing which item it used. Sourcesupport.smartbear.com/bearq/docs/en/context.htmlread 2026-11-20 |
|
| Human Oversight & Guardrails | Full |
|
Autonomy is configurable. In human directed mode a team triggers BearQ's actions and approves its proposals through approval gates before they are final, and proposed tests never join the regression suite until someone approves them. In autonomous mode scope limits, exclusion rules, time bounds and resource controls fence where the agents explore and what they execute. Sourcesmartbear.com/product/bearq/features/human-in-the-loop-controlread 2026-11-20 |
|
| Security, Identity & Governance | Full |
|
BearQ applies role based access controls, and its API keys are scoped to one workspace and visible only to workspace admins. SmartBear holds ISO/IEC 27001, lists GDPR, DORA and EU AI Act readiness, and publishes SOC 2 Type 2 reports for SwaggerHub, Zephyr, PactFlow and QMetry and a Type 1 for Reflect, although BearQ itself sits outside those report scopes for now. Sourcetrust.smartbear.comread 2026-11-20 |
|
| Observability & Auditability | Full |
|
Every agent action lands in an audit trail, in autonomous and human directed modes alike, and each workflow run is logged step by step. The API returns a task's full metadata and activity log and streams progress until it finishes, while daily summaries and compliance ready reports roll activity up. Each task also shows which piece of context the agent applied. Sourcesupport.smartbear.com/bearq/docs/en/bearq-api.htmlread 2026-11-20 |
|
| Memory & State Persistence | Partial |
|
Context persists for the workspace and carries into every later task: a global entry plus targeted knowledge items and restrictions that people write and can delete through the edit screen. The Application Model also carries forward between runs. BearQ names no memory the agents write for themselves with a stated lifetime, so what lasts is a store people maintain plus the map of the app. Sourcesupport.smartbear.com/bearq/docs/en/context.htmlread 2026-11-20 |
|
| Deployment & Data Residency | Not documented |
|
BearQ runs as a hosted SmartBear service on AWS, reached at its own API domain, while older SmartBear tools such as ReadyAPI and TestComplete install on premises. Hosting regions, a choice of region and any customer environment option for BearQ are unknown, since the trust center keeps its subprocessor list and product architecture behind an access request. Sourcetrust.smartbear.comread 2026-11-20 |
|
| Prebuilt Agents / Templates / Packs | Partial |
|
BearQ ships its three agents ready to work and comes with default workflows that teams run before building their own. The agents are roles inside one product that depend on each other, since the Tester runs what the QA Lead writes against the Explorer's map, and the default workflows lack the event triggers custom ones can take. SmartBear offers no catalog of separate agents to adopt one by one. Sourcesupport.smartbear.com/bearq/docs/en/bearq-agents.htmlread 2026-11-20 |
|
| Triggers & Channel Coverage | Full |
|
Workflows run on a configured schedule, from a CI/CD pipeline through the public API, from a coding agent through the MCP server or on demand, and custom workflows can also take event triggers once they have completed one run. BearQ explores the areas of the application that change with each release again, so testing keeps pace with deploys rather than waiting for someone to ask. Sourcesupport.smartbear.com/bearq/docs/en/workflows.htmlread 2026-11-20 |
|
| Model Flexibility & Routing | Not documented |
|
SmartBear runs generative AI across BearQ, Reflect, QMetry, Zephyr, TestComplete and Swagger without naming the models behind any of them or offering customers a choice. Reflect can author tests with the developer's preferred coding assistant, and the MCP server serves whichever assistant connects, but those are the customer's own models working through SmartBear tools rather than models BearQ runs. Sourcesmartbear.com/airead 2026-11-20 |
|
| APIs / SDKs / MCP Extensibility | Full |
|
The BearQ REST API creates tasks for the QA Lead, Tester or Explorer, starts and cancels saved workflows, and returns status, activity logs and streamed progress, authenticated by a workspace scoped bearer key with errors in problem JSON. The SmartBear MCP Server exposes test runs, BugSnag error context and Swagger specifications to AI assistants and IDEs, with install guides on SmartBear's developer site. Sourcesupport.smartbear.com/bearq/docs/en/bearq-api.htmlread 2026-11-20 |
|
| Testing, Debugging & Optimization | Partial |
|
BearQ tests the customer's application, and the Tester checks each journey against expected outcomes. What the agents themselves produce is reviewed through approval of proposed tests, activity history and reports, which let a team inspect and correct agent work after the fact. Nothing scores BearQ's exploration or test authoring against a benchmark or gates a change to its agents on results. Sourcesmartbear.com/product/bearq/features/human-in-the-loop-controlread 2026-11-20 |
|
| Browser / Computer-use | Full |
|
The Explorer navigates the customer's application on its own, moving through pages, elements and states to build its map, and the Tester walks end to end user journeys through the interface and checks what appears. That is the agent operating software it does not control, without a person steering each step. The browser technology underneath, hosted or otherwise, goes unnamed. Sourcesmartbear.com/product/bearqread 2026-11-20 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Pricing
Quote only, priced on agent compute units
Agent compute units consumed by BearQ agent work
Included quota
Every BearQ customer gets the same capabilities. The contract sets the agent compute units included.
What is public
SmartBear prices BearQ on agent compute units rather than seats or feature tiers, and more intensive agent work uses more units. A free seven day trial opens every capability with guided setup and usage insights.
Billing mechanics
Running, creating and investigating tests each consume agent compute units. Plans are built from expected usage, from one application with a few releases a month to many applications with high release velocity, and unused units roll over for the contract term.
Overage / add-ons
Unused agent compute units roll over for the life of the contract.
Sales call required
Yes, required for paid access
Free / trial
Free seven day trial
Commercial notes
Cost governance lets teams track usage, set alerts and limits and manage consumption across applications and teams. SmartBear's other products, such as TestComplete and ReadyAPI, sell through its online store.
Missing data
SmartBear quotes the unit price and the number of units in each plan after a conversation about expected usage.
Related vendors
- AgentOps — Agent observability and debugging platform: open source SDKs trace…
- AgentPhone — AgentPhone gives the AI agents a team builds their own US or…
- Agno — Python agent framework and AgentOS runtime (formerly Phidata) for…
- Airbyte — Airbyte moves data from operational systems into warehouses and…
- AIsa — Resource and payment gateway for AI agents, with one key to 110+…
- Albato — Albato is a no code integration platform that sells an embedded,…