Docs · CLI Reference (Commands & Flags)

CLI Reference (Commands & Flags).

Every flag that matters.

Direct answer

anv is keyboard-first with explicit approvals: 21 native tools (read/edit/patch/run/lsp/web/mind/subagents), 8 TUI screens, three run modes for three risk levels. This is the working reference; full help ships in the binary (anv --help).

CLI Reference (Commands & Flags)

Modes

01Interactive TUI · default8 screens: Startup, Agent, Git, Monitor, MCP browser, MindViz graph, Terminal panes, Swarm fleet. Keys 1-8, Tab, approvals inline (Y/N/A).
02--no-tui --yolo · CIHeadless scripted runs with exit codes 0/1/2 (denied). Pipeable I/O for pipelines and cron. Auto-approve only where revert is cheap.
03--plan-only · read-only Full exploration, zero mutations. Pair with repomap blast-radius for safe multi-file plans.
04anv init · anv modelsScaffold .anvaya/ (<50ms, idempotent); list and switch models/providers per project.

CLI Reference (Commands & Flags)

Tools (21)

01Read + navigateread_file (ranges), list_dir, dir_tree, glob, grep, search_in_file, lsp (pooled definitions, refs, diagnostics).
02Write + runedit_file, rewrite_file, create/delete, patch (unified diff), run_command, run_script, backups before mutation, sandbox-checked, approval-gated.
03Think + reachquestion (ask user), task (subagents), todowrite, mind (recall/record), webfetch, websearch, monitor (watch long-running processes).

Questions

Asked About CLI Reference (Commands & Flags).

Q

How do approvals work?

Inline ▸ APPROVE? cards per tool call: Yes / No / Always-this-session / numbered presets. Denials exit 2 in CI for clean gating.

Q

What happens on context pressure?

10-cell CTX meter (red-blink ≥90%), pre-flight gate at 95%, summarizer at 80% with hard-drop at 95%, plus Mind packing keeping turns small.

Q

Keyboard-first?

Fully: 1-8 screens, Tab focus, Ctrl+P palette, Ctrl+T theme, Ctrl+B transparency, V inspector, Space drill-down, F1 keys. Mouse optional.

Run Agents That Fit On Your Laptop.

25.6 MB median RSS. 25 agents ran in parallel on a Core 2 Duo with 4 GB RAM. Hundreds on your machine. Zero cloud required on the Ollama path.

Get Started →Read the FAQ

Requires Rust/cargo to build from source. Linux and macOS today, Windows not yet supported. Pre-1.0, public beta. Pricing TBD.