Blog / New tools

IDE test agents and unit test generators: creation, output and agent integration

Published by AI QA Directory on . Built from the catalogue data as of that date; tool pages carry the current facts.

The catalogue lists 14 IDE and repo test agents and 17 unit test generation tools, as of 2026-09-24. This post lists all 31 with three facts: how tests are created, what the output is, and how the tool connects to coding agents. Each tool page gives the source and verification date for every fact.

How to read the tables

Both tables are sorted by tool name, A to Z. Every cell is documented unless it says "vendor-stated" or "not documented". In the agent integration column, "not recorded" means the catalogue holds no agent integration fact for the tool.

IDE and repo test agents

The category definition reads: "Copilots, plugins and agents that write tests inside the repository."

IDE and repo test agents
ToolCreationOutputAgent integrationStatus
Agentic QE Fleetprompt to test, code generationnot documentedMCP server; Claude Code plugin with agents, slash commands and skillsactive
Amazon Q Developer unit test agentcode generationnot documentedIDE plugins for VS Code, JetBrains, Eclipse and Visual Studio; supports MCP servers as a clientactive
Axolotlnatural languagenot documentedVS Code extension built on Cline; writes a fix prompt for Claude Code, Cursor, Codex CLI or other agentsactive
Copilot for testsprompt to test, code generationnot documentedGitHub MCP Server; the Copilot cloud agent uses the Playwright MCP server by defaultactive
Cypress Copilotcode generationstandard Cypressnot recordedactive
Devin for testsprompt to test, code generationnot documentedofficial Devin MCP server; Devin connects to external MCP serversactive
Janknatural language (vendor-stated)not documentedpackages for Claude Desktop, Claude Code, Codex, Cursor and Antigravity plus an MCP server (vendor-stated)active (vendor-stated)
JetBrains AI Assistant test generationcode generationnot documentedMCP client; JetBrains IDEs from 2025.2 include an MCP server for external clientsactive
Playwright test agentsprompt to test, code generationstandard Playwrightagent definitions for VS Code, Claude Code, Codex and opencode, built on Playwright MCP toolsactive
Qodocode generation (vendor-stated)not documentedAgentic Toolbox MCP server, not for test generation; plugins for Claude Code, Codex and Kirorenamed
Qurenatural language, recording, code generation (vendor-stated)not documentednot documentedactive (vendor-stated)
Retraciorecording, code generation (vendor-stated)standard Playwright (vendor-stated)not documentedactive (vendor-stated)
Shiplightnatural language, recording, prompt to testproprietary runnerMCP server; /shiplight skill for coding agents; VS Code extensionactive
Tabnine test agentcode generation, prompt to testnot documentedconnects to MCP servers configured by the useracquired

Unit test generation

The category definition reads: "Tools that generate unit tests for existing code."

Unit test generation
ToolCreationOutputAgent integrationStatus
BitDiverecordingnot documentedMCP server exposing runtime traces and test tools to MCP clientsactive
ChatUniTestcode generationnot documentednot recordedactive
CodeRabbitcode generationnot documentedCodeRabbit CLI Claude Code plugin; CLI also documented for Cursor and Codexactive
CoverUpcode generationnot documentednot recordedactive
Diffblue Covercode generationnot documentedCover MCP Server (Beta) for MCP clientsactive
EarlyAIcode generationnot documentednot recordedshut down
GitAutocode generationnot documentednot recordedactive (vendor-stated)
GPT4Gocode generationnot documentednot recordedactive
Mutahunternot documentednot documentednot recordedactive
Parasoft Jtestcode generationnot documentedJtest MCP Server extension; agent skills for GitHub Copilot CLI and OpenAI Codex CLIactive
Pythagora (test generator)code generationnot documentednot documentedshut down
RoostGPTcode generation, natural language, autonomous from URLstandard PlaywrightRoostGPT VS Code extensionactive
Symflowercode generationnot documentednot documentedactive (vendor-stated)
Test Grucode generationnot documentednot documentedactive
TestPilotcode generationnot documentednot recordedshut down
TestSparkcode generationnot documentednot documentedactive
Tuskcode generation, recordingnot documentedTusk Drift MCP server for Cursor, Claude Code and Claude Desktop; agent skillsshut down

Counts across the 31 tools

  • Output is not documented for 26 of the 31 tools.
  • Standard Playwright output is documented for Playwright test agents and RoostGPT.
  • Standard Playwright output is vendor-stated for Retracio.
  • Standard Cypress output is documented for Cypress Copilot.
  • Proprietary runner output is documented for Shiplight.
  • The agent integration fact names MCP for 14 of the 31 tools. For Jank, that fact is vendor-stated.
  • Code generation is a documented creation method for 23 of the 31 tools.

Status changes

Six of the 31 tools are not listed as active.

  • EarlyAI: vendor docs list the IDE extension as a retired product with archived documentation.
  • Pythagora (test generator): a README commit marked the repository as deprecated on 2024-03-09.
  • Qodo: CodiumAI announced its new name, Qodo, on 2024-09-30.
  • Tabnine: Tricentis announced its acquisition of Tabnine, dated 2026-07-30.
  • TestPilot: GitHub Next archived the repository on 2025-02-03.
  • Tusk: Tusk announced that it sunset the product on 2026-06-10.

The AI coding agent workflow page lists tools across all categories whose agent integration fact names an MCP server, an IDE plugin or a coding agent. The IDE test agents and unit test generation hubs hold the full fact sets.

Sources

Spotted an outdated or wrong fact? Send the source and we apply it at the next monthly update.

Report a change