Gentle-AI
A Go CLI that configures the AI coding agents you already use: it gives them shared memory, a workflow and deterministic steps without replacing the agents them
Medium risk
We rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.
Why this level
- Installs a third-party binary and writes into your agents' configuration directories
- Reads the codebase for the symbol graph and has opt-out telemetry
Install
Manual install
brew install gentleman-programming/tap/gentle-aimacOS via Homebrew. After install run gentle-ai to pick agents and components.
This is third-party code. Review the repository files before installing.
What it does
Gentle-AI is a Go binary that adds a shared workflow to your agents through their native configuration mechanisms rather than its own shell. It brings Engram memory so the agent records decisions and reaches for them in new sessions, plus work modes: lightweight ODD for small edits, formal SDD with spec and design phases, and an optional RDD code review. The binary itself owns the SDD status and RDD review transitions and reads state from files on disk, so the next valid step is the same for different people and models. It also includes a skills library, an optional Context7 MCP for documentation, a read-only symbol graph of the codebase, a deny-list for secret files, config backups before every write, personas and themes. Sixteen agents are supported, among them Claude Code, Codex, Cursor, OpenCode and Gemini CLI, and Gentle-AI never installs an agent for you.
Who it is for. For developers and teams who want a single workflow and a shared quality bar on top of different AI agents.
Good fit when
- The team uses several different agents and needs a shared workflow and memory
- You want deterministic SDD steps and an optional code review on top of the agent
- You want the agent to keep project context between sessions
Not a fit when
- You use a single agent and its built-in setup is enough
- You cannot install a third-party binary or change agent configs on the machine
Example request
Set up my Claude Code and Cursor with gentle-ai: enable memory and the SDD workflowLimitations
Gentle-AI configures agents you already have and does not install them, so the available features depend on what each agent supports. Installation is via Homebrew, a curl script or go install, and building from source needs Go 1.25.10 or newer. The tool writes into the agents' configuration directories and snapshots them before each write. There is telemetry, which can be turned off.
How to disable. Uninstall the binary you installed via Homebrew, the script or go install. Agent configs can be restored from the backups Gentle-AI makes before every write.
Security check
- Installs a third-party binary and writes into your agents' configuration directories
- Reads the codebase for the symbol graph and has opt-out telemetry
README in short
The README presents Gentle-AI as a deterministic engineering environment on top of the agent you already use. Key components are Engram memory, the ODD, SDD and RDD review modes, and the gentle-ai binary that reads state from files and returns the single valid next step. It also documents a skills library, Context7 MCP, a symbol graph, a secrets deny-list, config backups, personas and themes, and support for 16 agents. Installation is via Homebrew, a curl script or go install, after which the gentle-ai command configures the chosen agents and gentle-ai doctor checks the setup. The project is MIT licensed, and the Gentle-AI and Engram names are the author's trademarks.
FAQ
Does Gentle-AI replace my agent?
No. It does not install or replace agents; it configures the ones you already have through their native configuration mechanisms.
Why is the next step deterministic?
The gentle-ai binary owns the SDD status and RDD review transitions and reads state from files on disk, so the answer about the next step does not depend on the model and is the same for different people.
Related
Open-source personal AI assistant on your own machine: answers in Telegram, Slack, Discord and WhatsApp, extended with skills and plugins
A self-improving agent from Nous Research with a TUI, messaging gateway, cron jobs and skills it writes itself
An open source coding agent for the terminal and desktop with build and plan modes
Open prompt library with a Claude Code plugin, MCP server and CLI: search, fetch and improve prompts and skills from an agent