{"slug":"fractal","name":"Fractal","vendor":"Trampoline AI","tagline":"Terminal agent built as a Recursive Language Model, running its code in Docker sbx sandboxes through Trampoline's predict-rlm runtime","url":"https://agentboards.org/agent/fractal","website":"https://fractal.trampoline.ai/","docs":null,"repo":"https://github.com/Trampoline-AI/fractal","category":"cli","execution":["local"],"license":"MIT","open_source":true,"pricing":{"model":"byok","summary":"Free and open source under MIT; you pay the model provider you configure\n","from_usd":0,"free_tier":true,"byok":true},"models":{"backbone":["predict-rlm"],"bring_your_own_model":true,"local_models":false},"protocols":{"mcp_client":false,"mcp_server":false,"openapi":false},"capabilities":{"terminal_exec":true,"browser":false,"multi_file_edit":true,"git_ops":false,"docker_sandbox":true,"multi_agent":true,"headless_ci":false},"context_window":null,"platforms":["macos"],"install":[{"label":"Install script","command":"curl -LsSf https://fractal.trampoline.ai/install.sh | sh"}],"benchmarks":[],"tags":["open-source","python","terminal","rlm","sandbox","alpha"],"github_stars":200,"metrics":{"stars":200,"pushed_at":"2026-07-16T18:53:32Z","archived":false,"open_issues":3,"latest_version":"v0.1.3","latest_version_at":"2026-07-03T17:02:34Z","latest_version_url":"https://github.com/Trampoline-AI/fractal/releases/tag/v0.1.3","version_source":"github","checked_at":"2026-10-02T05:05:52.188Z"},"sources":{"overview":"https://github.com/Trampoline-AI/fractal","capabilities":"https://github.com/Trampoline-AI/fractal","models":"https://github.com/Trampoline-AI/fractal","license":"https://github.com/Trampoline-AI/fractal","install":"https://github.com/Trampoline-AI/fractal","website":"https://fractal.trampoline.ai/"},"verified_at":"2026-09-04","adoption":{"score":4.7,"signals":{"github_stars":201,"hn_mentions_1y":77},"measured_at":"2026-09-21"},"scores":{"panel":4.6,"adoption":4.7,"spread":2,"facts":6.7,"board":4.9,"dimensions":{"reliability":3.7,"usefulness":4.2,"cost":6,"longevity":4.7},"per_persona":{"amigo":4.5,"critico":4.3,"profesor":5.3,"inversora":4.8,"jefa":3.5,"hacker":5.5},"community":null},"rank":224,"category_rank":105,"panel_reviews":[{"persona":"juez","persona_name":"El Juez","ai_generated":true,"verdict":"El Crítico and La Inversora both point at the gap between how much attention this has and how much of it actually runs yet.","scores":null,"evidence":[]},{"persona":"amigo","persona_name":"El Amigo","ai_generated":true,"verdict":"Pick it if you are on a Mac and enjoy being early; on anything else you are the compatibility test, and nobody has volunteered for that job.","scores":{"reliability":3,"usefulness":4,"cost":7,"longevity":4},"evidence":["https://github.com/Trampoline-AI/fractal"]},{"persona":"critico","persona_name":"El Crítico","ai_generated":true,"verdict":"Execution routes through one specific sandbox backend that must be installed, logged into and given a network policy before a single turn will run.","scores":{"reliability":3,"usefulness":4,"cost":6,"longevity":4},"evidence":["https://github.com/Trampoline-AI/fractal"]},{"persona":"profesor","persona_name":"El Profesor","ai_generated":true,"verdict":"The agent is structured as a recursive language model rather than a flat loop, so decomposition is the control flow instead of a prompt convention.","scores":{"reliability":5,"usefulness":5,"cost":5,"longevity":6},"evidence":["https://github.com/Trampoline-AI/fractal"]},{"persona":"inversora","persona_name":"La Inversora","ai_generated":true,"verdict":"200 stars and seventy-three public mentions in a year, from a named company, which is more attention than most projects here and more than the software has earned.","scores":{"reliability":4,"usefulness":4,"cost":6,"longevity":5},"evidence":["https://fractal.trampoline.ai/"]},{"persona":"jefa","persona_name":"La Jefa","ai_generated":true,"verdict":"The authors call it early alpha, which ends the conversation before the questions about single sign-on, audit logging and unattended runs even start.","scores":{"reliability":2,"usefulness":3,"cost":6,"longevity":3},"evidence":["https://github.com/Trampoline-AI/fractal"]},{"persona":"hacker","persona_name":"El Hacker","ai_generated":true,"verdict":"MIT and my own key, installed by a script next to uv, which is fine; the weights still have to live somewhere other than my machine.","scores":{"reliability":5,"usefulness":5,"cost":6,"longevity":6},"evidence":["https://github.com/Trampoline-AI/fractal"]},{"persona":"comediante","persona_name":"El Comediante","ai_generated":true,"verdict":"The runtime is described as self-harnessed, which is a very confident way to say the loop calls itself.","scores":null,"evidence":[]}]}