gentle-ai
Gentle-AI is an ecosystem configurator. It takes the AI coding agent runtime already on your machine (Claude Code, OpenCode, Cursor, Codex and others) and equips it with persistent memory, a planning workflow, a curated skill library, MCP tool servers and model routing. The README's pitch is that an agent that normally forgets everything between sessions can instead remember decisions, follow your project's conventions and produce reviewable evidence for what it changed. It never installs an agent for you; if you select one it cannot detect, it refuses and prints the command to run yourself.
You choose components individually or through presets. They include Engram (persistent memory across sessions), Skills, an optional teaching-style Persona, Spec-Driven Development for larger features, Context7 for live library documentation, Permissions with a deny list for ~/.ssh, .env and credential files, a provider switcher and themes. Presets range from Dev Stack and Polish (everything) to Memory Only. Different agents are configured through their own native features, so capabilities differ: Claude Code gets sub-agents and output styles, while OpenCode gets per-phase model routing. Start by running `gentle-ai`, then verify with `gentle-ai doctor`. The project is MIT licensed.
Features
Why choose it
Trade-offs
Maintenance check
- Last commit
- 2026-11-11
- Commits, last 90 days
- 100+
- Releases, last 90 days
- 30
- Latest release
- v4.4.0 (2026-11-07)
- Open issues
- 409
- Recent issues with at least one comment
- 7/30
Oldest week on the left; the latest four weeks are orange. Source: GitHub commit activity.
Status rule: archived on GitHub = Archived; last commit within 90 days = Active; 91 to 365 days = Slowing; older = Stale. All figures come from the GitHub API on the date shown.
Most discussed open issues
- #4571 bug(review 2.8.2): correction_required capture's own status_continuation classifies the just-minted authority corrupted_or_unverifiable_auth393 comments · 2026-09-14
- #5422 [Automated provider defect] bound STATUS after granted START times out pre_native for ~800+ changed-line candidates on 4.0.0 stable (Windows145 comments · 2026-10-09
- #4655 bug(review): bound STATUS still times out pre_native after a granted START on a 3-path medium candidate (2.9.1 stable) — regression of #3733106 comments · 2026-09-16
- #5129 Review integration blocks on undocumented intended-untracked capture66 comments · 2026-09-30
- #5209 [Automated provider defect] Claude Code: a granted, never-reoffered START's first capture-result lands directly on correction_required, and 58 comments · 2026-10-02
Issue titles are quoted from GitHub. Open the link to check whether a problem still applies to the version you use.
Who it fits
- You want to add persistent memory, skills, a planning workflow and MCP servers to an agent you already run.
- You want to pick components one by one or use presets.
- You want a deny list for ~/.ssh, .env and credential files.
- Your agent does not support sub-agents: the README says some capabilities differ by runtime.
- You only want one piece such as memory and prefer a single focused tool.
- You do not want stylistic extras such as personas and themes in your setup.
How it differs
ZCF focuses on one-command setup of workflows, API settings and MCP for Claude Code and Codex. Gentle-AI layers memory, skills and permissions onto several agents.
Compatibility
Quick start
Use cases
Alternatives
Featured in these lists
Comments
No comments yet. Be the first!