Total recall for
your terminal.
Search your shell history. Track AI agent commands.
Keep command context in a local SQLite database. Connect supported AI agents through MCP to query past work.
Already use Homebrew?
brew tap AppachiTech/suvadu && brew install suvaduThen enable shell integration below. Install script and Cargo options are there too.
Using AI coding agents? See a session handoff or inspect captured sessions.
Recorded with fictional data in v0.5.0 · Search guide
Not just history. Memory.
Three layers that turn your terminal history into a structured, queryable, agent-accessible knowledge base.
Structured History
Recorded commands include exit code, duration, directory, session, and executor. Search and filter your history in a terminal UI.
AI Agent Tracking
Identify supported agents in recorded commands. Configure an agent integration to capture prompts and link them to commands where supported.
MCP Agent Memory
Query history and past sessions through MCP, and share reusable instructions with the new skills library. Sync skills into Claude Code, Cursor, and Codex.
Suvadu stores history locally. When you connect an AI client, that client controls how retrieved history is used.
See it in action
Discover features in Home and follow an agent session from prompt to commands. Try command recall in the Search demo above.
Hand off an agent session
A Claude Code session stops partway through a task. Suvadu has already imported its prompts, responses, and commands. Open Codex in the same repo and ask it to continue; Codex reads the session over MCP.
$ suv sessions --list -n 2
Last active ID Type Tag Agent / Host Model Tokens Cmds Duration
────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
2026-09-19 09:14 claude-8 AI — claude-code claude-sonnet-4-5 47330 9 42.0m
2026-09-19 08:55 4f2a9c1e Human api laptop — — 31 72.0myou › Pick up where my Claude Code session left off in this repo.
called list_agent_sessions
called get_agent_session claude-8f3c2a71-…
codex › The last Claude Code session added rate limiting to the login route. Its final npm test run (command-5512) exited 1: two tests in tests/auth.test.ts still expect 200 instead of 429. I'll update those tests first.
Synthetic example. Claude Code, Codex, and OpenCode sessions are imported natively. Other integrations record commands and prompts. A turn still in progress is available after it ends.
Discover features and understand AI activity
Recorded in Suvadu 0.5.0 with fictional history, prompts, and responses. No live agent work is performed. See the CLI docs for full instructions and shortcuts.
Find the feature you need
Run suv to explore tools by task, or type a feature name and open it directly.
- Browse categories without reading the entire manual
- Search for a task such as “failed”
- See the command and shortcut behind each feature
- Open filtered history and return to Home with Esc
What happens in this video
The demo runs suv, opens Find a command, types failed, and presses Enter to open history filtered to failed commands. Esc returns to Home.
Why switch?
Add structured context and filtering to your shell history.
| Feature | Standard History | Suvadu |
|---|---|---|
| Search | Built-in history search | Substring search with interactive filters |
| Command context | Commands; timestamps depend on configuration | Exit code, duration, directory, session and executor |
| Arrow recall | Usually recent commands first | Zsh: recent commands first, directory as a tiebreaker |
| Filtering | Shell commands and external tools | Date, exit code, directory and executor filters |
| Storage | A shell history file | A local SQLite database |
| Analytics | External tools or scripts | Built-in stats and agent dashboards |
| Recording controls | Shell-specific ignore settings | Ignore patterns, pause and disable controls |
| Secret redaction | Depends on shell and tooling | Pattern-based redaction for recognized secrets |
Shell behavior varies by configuration. Suvadu records commands after integration is enabled; ignore settings apply. Redaction matches known patterns and cannot identify every secret.
Works with your tools
Dedicated integrations and executor detection for supported tools.
Shell recording, prompt capture and MCP access have different setup requirements. Detection alone does not capture prompts. Choose your agent setup or see the compatibility matrix.
All trademarks belong to their respective owners.
Install Suvadu
macOS & Linux · Zsh 5.1+ or Bash 4+
Install the CLI, enable recording in your shell, then try a search.
-
Install the CLI
Choose one method. Each installs the
suvcommand.Requirements and manual installationHomebrew
Recommended if you already use Homebrew on macOS or Linux. Update later with brew upgrade suvadu.
brew tap AppachiTech/suvadu && brew install suvaduInstall script
Works on macOS and Linux, Intel and ARM. Downloads a binary for your operating system and architecture, verifies its checksum, and installs to /usr/local/bin, using sudo only if that directory is not writable. End the command with bash -s -- --user instead to install into ~/.local/bin without sudo.
Read the installer sourcecurl -fsSL https://downloads.appachi.tech/suvadu/install.sh | bashCargo
Requires a Rust toolchain. Compiles from source, which can take a few minutes. Make sure ~/.cargo/bin is on your PATH.
cargo install suvadu -
Enable shell integration
Installing the CLI alone does not record commands. Choose the shell you use to enable recording and Ctrl+R search.
Zsh setup
Add this line once to
~/.zshrc. If it is already there, keep the existing line.eval "$(suv init zsh)"Save the file, then run this in your Zsh terminal to activate it:
source ~/.zshrcBash setup
Add this line once to
~/.bashrc. If it is already there, keep the existing line.eval "$(suv init bash)"Save the file, then run this in your Bash terminal to activate it:
source ~/.bashrcBash 4+ is required. For a login shell, make sure
~/.bash_profilesources~/.bashrc, or add the hook to the startup file your shell reads. Shell setup details -
Check your setup and try a search
Check your shell, recording state and database. Agent and MCP setup are optional; follow the diagnostic guidance for any issues.
suv doctorRun
echo hello, then press Ctrl + R and search for it. Select a result to place it in your prompt; press Enter to run it.Need help? Read the diagnostic guide or check shell integration.