AgentRecall

A unified lens to search, inspect, and export conversation history across AI coding agents in VS Code.

🌟 Supported Agents

Agent	Status	Supported Path / Format
OpenAI Codex	✅ Supported	`~/.codex/sessions/` (JSONL rollouts & threads)
OpenCode	🔜 Planned	`~/.local/share/opencode/`
AGY (Antigravity)	🔜 Planned	`~/.gemini/antigravity-cli/brain/`
Claude Code	🔜 Planned	`~/.claude/sessions/`

🔍 Real-Time Full-Text Search (Ctrl + Alt + F / Cmd + Alt + F)
- Search across User Inputs, Agent Thinking Processes, and Assistant Responses.
- Clearly tags match categories ([User], [Thought], [Response]) with contextual previews.
🎯 Precise Jump & Highlight
- Selecting any search result immediately opens the conversation document, scrolls the editor directly to the matching line, and highlights the keyword.
🧼 Boilerplate-Free Clean Markdown
- Automatically filters internal framework prompts, AGENTS.md injected templates, and noisy empty tool call frames.
- Generates clear, readable sections:
  - 👤 ## 👤 User
  - 💭 > 💭 Thought / Preamble
  - 🤖 ## 🤖 Response
📂 Dual Sidebar Views
- Browse historical sessions by update time in both the Activity Bar (AgentRecall) and the Explorer (File Tree) panel.
🚀 Workspace Sync for Native Ctrl + Shift + F
- Export all sessions as clean .md files into .agentrecall-history/ in your workspace, allowing VS Code's native global search to index all agent dialogues.

Shortcut / Command	Action
`Ctrl + Alt + F` (Mac: `Cmd + Alt + F`)	Search Agent History: Global real-time dialogue search
`AgentRecall: Export All Chats to Workspace`	Export all sessions to workspace for native global search
`AgentRecall: Refresh History`	Reload and refresh the conversation list

Built as a lightweight, zero-dependency VS Code extension.
Uses VS Code's native TextDocumentContentProvider (agentrecall:/chat/<id>.md) for memory-efficient virtual Markdown rendering.
Extensible multi-agent parser architecture designed to support disparate local session formats seamlessly.