Save versions of your AI prompts, test changes before shipping, and roll back if something breaks. Works completely offline with your local CLI or synced to the cloud.
Open source • Works offline with SQLite • Your keys stay on your machine
Manage, version, diff, and evaluate atomic prompt bundles.
| Prompt Name | Description | Version | Test Suite | Visibility | Actions |
|---|---|---|---|---|---|
| security-audit-v2 | Automated vulnerability scanning and CWE classification for Node & Python apps. | v3 | 12/12 (100%) | Private | Open studioEdit bundleView diffRun evals |
| customer-support-agent | Tone-calibrated customer response generator with strict refund guardrails. | v5 | 23/24 (96%) | Public | Open studioEdit bundleView diffRun evals |
| code-reviewer-pro | Strict TypeScript and Rust code review agent with security audit rules. | v4 | 18/18 (100%) | Private | Open studioEdit bundleView diffRun evals |
When prompts are hardcoded in application files, small edits break formatting with no way to undo bad changes. Git for Prompts gives you clear commit history, side-by-side diffs, and test suites.
Prompt edits pushed straight to code without test runs, schema checks, or rollback points.
Snapshots saving your prompt template, model choice, and output format in a local SQLite file.
From initial draft to automated evals and team sync.
Initialize a local SQLite repository directly inside your project folder.
Run local evaluations with your environment API keys against real test cases.
Sync approved versions to the team dashboard or deploy over the REST API.
Run test cases against dual AI models (Groq & OpenRouter). Validate variable interpolations and strict Zod JSON output schemas.
Input: SQL Injection payload • Score: 100%
Zod validation: Object schema verified
Interpolated "{{codeSnippet}}" successfully
Version control your prompts, model settings, and structured outputs without cloud lock-in.
Every save creates a new version with full change history. Roll back instantly or view changes side-by-side.
The CLI runs completely offline in your terminal using local SQLite. No cloud account required.
Keep your system prompt, model settings, and response schemas saved together in one place.
Push prompt versions to the web dashboard and serve them to your apps using secure API keys.
Run evaluations, inspect diffs, and push approved prompt versions directly from your terminal. Pure TypeScript shared between the CLI and web dashboard.
Explore CLI Docs →$ gitforprompts diff prompt-v1 prompt-v2
--- prompt-v1 (2026-08-06)
+++ prompt-v2 (2026-08-07)
- systemPrompt: "You are a helper."
+ systemPrompt: "You are a strict code reviewer."
+ tools: ["run_static_analysis", "lint_check"]
“Treat prompts like production code: version them, diff them, and test them in your terminal before they hit production.”
Prompts control how AI apps behave, but most teams still keep them scattered across code files. Git for Prompts brings git-like version control to your prompts: clean version history, side-by-side diffs, automated testing, and local storage.
You never need an internet connection. The CLI saves prompts in a local SQLite database inside your project (.gitforprompts/). Create prompts, save versions, and run tests completely offline.
When syncing prompt versions across a team, database locks ensure version numbers never collide or overwrite someone else work.
We never store your AI API keys. Tests run using your own local environment variables, so your secrets never leave your machine.
Built for both human developers and AI coding assistants. Supports clean schemas, standard API endpoints, and direct CLI commands.
Built in public. We welcome contributions from developers, researchers, and AI engineers.
Everything you need to know about Git for Prompts.
No. Local evaluations run in your terminal using your local environment variables. On the cloud platform, API credentials are stored as non-reversible SHA-256 hashes.
Run `$ npx gitforprompts init` in your project folder. Complete history, works offline, zero setup.
Get started free →