⌁
Coverage-aware agents
Starts from real coverage output, then focuses agent effort on gaps that can improve the suite.
Local-first AI-assisted test generation
Grow your test suite until coverage finally catches up.
AgenticTesting measures your repository's coverage, finds meaningful test gaps, plans the next focused task, and lets AI agents improve tests through constrained local tools.
5-stepagentic coverage loop
Localsettings and secrets stay on device
Multi-stackPython, .NET, C++, Jest, custom
coverageCurrent line coverage: 62%
analyzerDetected under-tested parser and auth paths
plannerNext task: add focused tests for edge cases
agentEditing test files through safe workspace tools
reviewerValidating result and preparing next iteration
targetLoop continues until requested coverage is reached
The quintessence: less manual test hunting, more guided coverage growth.
Instead of staring at coverage reports and guessing where to write the next test, AgenticTesting turns the process into a repeatable local workflow.
⌘
Local browser dashboard
Pick a repository, set a target, choose a provider/model, inspect runs, and reuse saved settings.
◈
Safer repo operations
Agents operate through relative file tools and command allowlists rather than unrestricted shell access.
A simple loop for a hard problem.
The product is built around one clear cycle: measure, analyze, plan, implement, review, repeat.
1
Measure
Run the configured coverage command for the selected project.
2
Analyze
Find files, branches, or behavior that need useful tests.
3
Plan
Select one focused testing task instead of trying to fix everything.
4
Implement
Let the agent add or improve tests through constrained tools.
5
Review
Check the result and continue until the target or limit is reached.
Run it locally.
The primary workflow is the local UI. API keys can be saved encrypted locally, and local model options are supported for laptop-first usage.
git clone https://github.com/sl-badcoder/AgenticTesting.git
cd AgenticTesting
python -m venv .venv
source .venv/bin/activate
pip install pytest pytest-cov cryptography
python -m src.frontend.debug_ui
# open http://127.0.0.1:8765Testing should not stop at “the build passed.”
AgenticTesting gives AI a narrow, practical job: inspect coverage, create targeted tests, and keep iterating with guardrails.
Explore the repository