Before you start
- ✓A project with Git initialized
- ✓Node.js 18 or later
- ✓A AIFlare account·Don't have an account yet?
- ◆Claude Code CLIInstallation guide ↗
- ◆OpenAI Codex CLIInstallation guide ↗
- ◆Gemini CLIInstallation guide ↗
- ◆GitHub Copilot CLIInstallation guide ↗
API key setup
Sign up at aiflare.dev, then go to Project Settings → API Key Management and generate an API Key. Place the downloaded aiflare.yml at your project root.
Run the installer
With aiflare.yml in your project root, run the command for your OS. The installer asks which agents you use (Claude Code, Codex, Gemini CLI, Copilot CLI — one or more), then wires up the matching skills, hooks, MCP server, and a Git pre-push hook for each.
macOS / Linux
curl -fsSL https://aiflare.dev/install.js -o install.cjs && node install.cjs && rm install.cjs
Windows (PowerShell, with curl.exe — Windows 10 build 1803+)
curl.exe -fsSL https://aiflare.dev/install.js -o install.cjs; node install.cjs; del install.cjs
Windows (PowerShell, with Invoke-WebRequest — any PowerShell 3.0+)
iwr -useb https://aiflare.dev/install.js -OutFile install.cjs; node install.cjs; del install.cjs
Check the installed agent's skill list for context-capture (in Claude Code and Codex, type /skills).
> /skills Project skills (.claude/skills) context-capture
If not listed, check that .claude/skills/context-capture/ (Claude Code) or .agents/skills/context-capture/ (Codex, Gemini CLI, Copilot CLI) exists.
First capture
Ask your AI agent to do a simple task and commit — that's all it takes.
AIFlare only works when an AI agent commits. Manual git commits won't be captured.
Add a one-line project description to README.md and commit it
✓ If a timeline entry appears on the dashboard, you're all set!
Explore the timeline
Captured entries appear chronologically on the dashboard timeline. Each entry automatically records the commit's intent, change summary, and category.
Added a project introduction at the top of README.md per user request
featureExtracted JWT validation logic into a separate module for testability
refactorClick an entry on the timeline to view its details.
Entry detail — related prompts
Click an entry on the timeline to open the detail modal. View the commit's intent, change summary, and changed files, plus the related prompts (conversation history) in the right panel.
Added a description of the project's purpose and key features at the top of README.md per user request.
Added a project introduction section at the top of README.md and organized the key feature list
Add a one-line project description to README.md and commit it
I'll add a project description to README.md and commit it.
💡 The prompt panel lets you see the user-AI conversation, so you can quickly understand the context behind changes without a code review.
Work sessions
Select a work session from the sidebar to see its commits. Use the "View Prompts" button to view the full conversation for the session.
Clicking "View Prompts" opens the session prompt modal showing the full user-AI conversation for this session.
💡 Work sessions let you see the full journey toward a goal — every step and every prompt in context.
Project settings
Create/delete API keys, manage members, and edit project info from Project Settings.
Having trouble?
I'm getting an API Key error
Make sure aiflare.yml exists in your project root and that api_key and endpoint are correctly set.
Skill is not recognized
Verify that the .claude/skills/context-capture/ directory (Claude Code) or .agents/skills/context-capture/ directory (Codex, Gemini CLI, Copilot CLI) exists in your project root.
Capture doesn't appear on the dashboard
Make sure your AI agent is the one committing. Also check that the endpoint in aiflare.yml is correct.
Can I capture manual commits?
Currently, only AI agent commits are captured. The Skill analyzes the agent's conversation context, which isn't available for manual commits.
Curious how AIFlare works under the hood? How It Works →