Mininglamp Open-Sources Mano-AFK, an AI Agent That Codes, Tests, and Deploys Autonomously
Mininglamp has released Mano-AFK, an open-source autonomous execution framework under the Apache 2.0 license, designed to take a natural language prompt and carry out the full software build-test-deploy loop without human intervention. The system generates a product requirements document, writes code, deploys it locally, and runs three layers of testing — lint checks, API tests, and browser-based end-to-end tests — automatically looping back to fix errors it encounters. A separate adversarial review agent also inspects the final output for issues such as missing input validation. For browser-based testing, the tool can use either a cloud-based Claude model or a local 4-billion-parameter vision-language model optimized for Apple Silicon, which achieved 58% accuracy across 100 test cases on five web applications. The team noted that while the fix loop resolved around 70% of basic code errors autonomously, the local model struggled with dynamic UIs, animations, and complex multi-step user flows.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.
Discussion (0)
Log in to join the discussion and vote.
Log in