Open the persistent console, keep talking, and say “try again” to resume a saved task.
Open source · Local first · BYOK
Local intelligence, under your control.
LOLM is a terminal-native agent that can answer, code, run checks, and create files on your computer. Bring any model. Its own NFET system measures the trajectory and decides when to retrieve, verify, branch, or finish.
No browser agent. No LOLM account. No hosted execution.
● Ollama local · qwen3:14b ◆ NFET active Talk normally. I remember this session. YOU Make me a PDF launch plan and put it on my Desktop LOLM Researching and writing your document… ◆ Checking the result Created /Users/you/Desktop/lolm-document.pdf YOU _
Use your existing provider, Ollama, or any OpenAI-compatible endpoint.
Files and commands run locally. Remote providers receive only the inference context you choose to send.
NFET controls the loop.
The provider is the voice. LOLM's frozen local backbone and trained graft re-read candidate output and compute real per-token measurements. The calibrated controller acts on them.
stateₜ = [ corrected_hidden, entropy, drift, gate, regime ]
control → continue · retrieve · verify · branch · finalize
control → continue · retrieve · verify · branch · finalize
Use the model you already trust.
OpenAIAnthropicGeminixAIOpenRouterGroqMistralDeepSeekTogetherCerebrasOllamaCustom