Minima CLI
minima is a terminal coding agent. It reads and edits your code, runs commands, and answers questions about a repo — like any agentic CLI — but every prompt is first sent to the Minima recommender to pick the cheapest model that will actually do the job. It then runs that model itself (calling the provider API directly), grades the outcome, and reports feedback so the next pick is sharper.
recommend ──▶ run the model yourself ──▶ judge quality ──▶ feedbackThe harness talks to the Minima recommender service over the same /v1/* HTTP contract as the SDK. It never proxies LLM calls — it calls OpenAI / Anthropic / Google / etc. directly with your provider keys, so your prompts and code never pass through Minima.
What you get
Every turn is routed to the cheapest model expected to clear the task, across OpenAI, Anthropic, Google, DeepSeek, xAI, Groq, and OpenRouter. Pin a model with /model or --model any time.
read, write, edit, bash, ls, glob, grep, web_fetch, and todowrite — gated by a permission system — plus task, which delegates subtasks to parallel, cost-routed sub-agents (watch them live with /tree).
Plan verification: plan steps carry shell checks that must go red→green before the agent may call them done, a confidence tier (🟢/🟡/🔴) on every verdict, and a recovery ladder that re-routes or replans on verified failure.
--budget 5 reserves before every turn, reconciles to realized cost, warns at 50/75/90/100%, and in enforce mode actually stops the run.
minima auth provisions a Mubit project scoped to the current repo, so each project's task → model → outcome history stays isolated — plus a curated local memory (/memory) projected into every turn.
Append-only sessions with /resume, /compact, git-shadow checkpoints, /undo, and /rewind.
How it fits together
- Interactive TUI — the default. A conversation view, status bar (model, cost, context), model picker, and slash commands. See Interactive TUI.
- Non-interactive —
minima -p "…"prints one answer;minima --mode json "…"streams machine-readable events for scripting. See CLI usage. - The recommender — the Python FastAPI service that ranks models. The harness needs only a
MUBIT_API_KEY(routing + memory auth) plus one provider key to call models. See Configuration.
Next steps
- Install the harness and run
minima auth. - Configure credentials and pick your providers.
- Walk the usage flows — the major features, what they do, and how to benefit from them.
- Learn the interactive TUI — keybindings and slash commands.
- Curious how it works inside? Under the hood — the turn lifecycle, verification gates, budgets, and memory.