An AI coding agent that lives in your terminal. Pick a model tier, write a prompt, and let it handle the rest. Script it, pipe it, alias it.
A full coding agent, not a chatbot.
The agent reads files, writes code, runs commands, and iterates. A local tool-calling loop means faster iteration with no round-trips for file reads.
Monkey Models work with no key. Or connect OpenRouter, Anthropic, sign in with ChatGPT/Codex via OpenAI OAuth, or use a local LM Studio fallback. No lock-in.
Pass screenshots and images to the agent. It can analyze UI, read error messages from screenshots, and understand visual context.
Drop markdown files to customize agent behavior. Use BANANA.md for project-specific context the agent reads automatically.
Pipe prompts in, chain with other tools, run in CI/CD. The CLI agent is a first-class terminal citizen. banana "fix all lint errors" --yolo
LM Studio fallback means you can code with a local model when you have no internet. Cloud models when you do. Your choice.
Pick the mode that matches your confidence level.
See what the agent will do before it does it. Review the plan, approve or adjust, then execute.
Ask questions about your codebase. The agent reads files and answers without making changes.
Let the agent code and auto-apply changes. No review, no confirmation. For when you trust it.
Three branded AI tiers tuned for coding. Match the monkey to the task. No model selection paralysis.
Goes all out. Complex refactors, architecture decisions, heavy lifting. Your technical co-founder.
Sharp and balanced. Feature implementation, bug fixes, code review. Gets it done right.
Surgical and fast. Quick fixes, inline suggestions, boilerplate. Code only, zero fluff.
The CLI is almost ready. Sign up to get access the moment it drops.
No credit card required. Free during beta.
We'll notify you the moment the CLI is ready.