auto-yes 0.16.0

Automatically respond 'yes' to interactive CLI prompts via PTY proxy

auto-yes

Auto-respond to AI CLI prompts. Never get blocked by interactive yes/no dialogs again.

auto-yes wraps any CLI command in a PTY (pseudo-terminal) proxy, watches the output stream for interactive prompts and automatically injects the correct response — so your overnight AI agent loops, CI scripts, and automation pipelines never block waiting for a human.

Works out-of-the-box with Claude Code, Cursor Agent, Gemini CLI, OpenAI Codex, GitHub Copilot, Amazon Q, Aider, and 10+ more AI CLI tools. Also handles generic [y/n], Continue?, SSH fingerprint, Terraform yes, and any custom regex you define.

╭─ without auto-yes ─────────────────────────────────────────╮
│  claude "refactor auth module"                              │
│  > Trust this folder?  1. Yes  2. No   ← blocks forever   │
╰─────────────────────────────────────────────────────────────╯

╭─ with auto-yes ────────────────────────────────────────────╮
│  auto-yes claude "refactor auth module"                     │
│  > Trust this folder?  1. Yes  2. No                       │
│  [auto-yes] responded 'y' (matched: ❯ 1. Yes, I trust…)   │
│  ✓ Continuing automatically…                               │
╰─────────────────────────────────────────────────────────────╯

---

Installation

pip install auto-yes

Windows users — for full PTY support add the optional extra:

pip install auto-yes[windows]

---

Quick start

Wrap an AI CLI tool directly (recommended)

auto-yes claude "fix the tests"          # wraps: claude "fix the tests"
auto-yes cursor chat "add unit tests"    # wraps: agent chat "add unit tests"
auto-yes gemini "review this PR"         # wraps: gemini "review this PR"
auto-yes codex "explain this function"   # wraps: codex "explain this function"
auto-yes aider --model gpt-4o           # wraps: aider --model gpt-4o
auto-yes copilot                         # wraps: gh copilot
auto-yes amazonq chat "help me"          # wraps: q chat "help me"

The profile name maps to the real binary automatically — run auto-yes list to see all profiles.

Overnight / automation mode (--auto)

For unattended overnight runs in corporate environments where --dangerously-skip-permissions is restricted:

# load ALL AI profiles, near-zero cooldown (0.05s), verbose logging
auto-yes run --auto -- claude --dangerously-skip-permissions "work through the backlog"

# or pipe through a loop
for task in tasks/*.md; do
    auto-yes run --auto -- claude "$(cat $task)"
done

--auto automatically:

• Loads patterns for all supported AI CLI tools
• Sets cooldown to 0.05s (fast batch mode)
• Enables verbose logging so you can audit what was approved

Heartbeat / keepalive (--heartbeat)

Send Enter automatically every N seconds of inactivity to keep Claude / Cursor Agent responsive when it is waiting for confirmation but no prompt pattern was matched:

# Shell mode: send Enter every 30s while idle
auto-yes --on --heartbeat 30

# Wrap mode: same, wrapping claude directly
auto-yes claude --heartbeat 30 "refactor the auth module"

# Combine with --auto for unattended overnight runs
auto-yes run --auto --heartbeat 30 -- claude "work through the backlog"

The timer resets on every real keystroke, so it never fires while you are actively typing.

Shell session mode

Start a persistent shell where every prompt is auto-answered:

auto-yes --on --cli claude
# → now inside an auto-yes session; type normally, prompts are handled
exit

Single-command mode

auto-yes run -- apt install nginx
auto-yes run --cli claude --cooldown 0.2 -- claude "fix tests"
auto-yes run --cli all -- ./deploy.sh   # match every AI tool + generic

Interactive shell session (shell alias)

auto-yes shell --cli claude --cli generic   # same as --on, more intuitive

Human-in-loop session control

Control a running auto-yes from another terminal — pause mid-run for manual interaction, then resume:

# Terminal 1: running auto-yes
auto-yes run --auto -- claude "big refactor"
# [auto-yes] PID: 12345  toggle: auto-yes toggle 12345

# Terminal 2: you want to take over for a moment
auto-yes ls                   # see running sessions and PIDs
auto-yes toggle 12345         # pause (prompts pass through to you)
# ... interact manually ...
auto-yes toggle 12345         # resume auto-responses

# Or use pause/continue explicitly:
auto-yes pause 12345
auto-yes continue 12345
auto-yes pause all            # pause every session
auto-yes continue all         # resume all

Status

auto-yes status            # check if active in current shell
auto-yes ls                # list all running sessions with PID, profile, uptime

---

How it works

┌──────────────────────────────────────────────────────────────────┐
│  Your terminal                                                   │
│       │  ↑                                                       │
│       ↓  │                                                       │
│  ┌─────────────────────────────────────────┐                    │
│  │  auto-yes  PTY  proxy                   │                    │
│  │                                          │                    │
│  │  1. fork child on PTY slave             │                    │
│  │  2. select() on master + stdin          │                    │
│  │  3. expand ESC[nC → spaces             │                    │
│  │  4. strip ANSI, resolve \r\r\n         │                    │
│  │  5. scan last N lines with regex       │                    │
│  │  6. inject response via master fd      │                    │
│  │  7. forward SIGWINCH / SIGTERM         │                    │
│  └─────────────────────────────────────────┘                    │
│       │  ↑                                                       │
│       ↓  │                                                       │
│  Child process (claude / agent / aider / …)                     │
└──────────────────────────────────────────────────────────────────┘

Anti-false-positive safeguards:

Guard	What it prevents
Typing suppress (1 s)	User's own keystrokes triggering auto-responses
Match cooldown (0.05 s)	Terminal redraws re-triggering the same prompt
Response cooldown (0.5 s default)	Rapid-fire duplicate responses
Bottom-up line scan	Matching stale text buried in scrollback
Buffer clear on input	Stale prompt text re-matching after user interaction

---

CLI reference

auto-yes <profile> [ARGS...]       wrap an AI CLI tool directly (recommended)
auto-yes --on [OPTIONS]            start an auto-yes shell session
auto-yes run [OPTIONS] -- CMD...   run a single command with auto-yes
auto-yes list, -l, --list          list all available CLI profiles
auto-yes patterns [CATEGORY...]    list prompt patterns (optionally filtered)
auto-yes add-pattern PATTERN       persist a custom regex pattern
auto-yes del-pattern PATTERN       remove a custom pattern
auto-yes status                    check if auto-yes is active in current shell
auto-yes ls                        list all running auto-yes sessions with PID
auto-yes pause <pid|all>           pause auto-responses for a session
auto-yes continue <pid|all>        resume auto-responses for a session

Options (for --on and run)

Flag	Description	Default
--auto	Overnight mode: all profiles, 0.05s cooldown, verbose	off
--response TEXT	Text to send when a prompt is detected	y
--cooldown FLOAT	Seconds between auto-responses	0.5
--delay FLOAT	Seconds to wait before approving; press any key to cancel	0
--heartbeat SECONDS	Send Enter every N seconds of inactivity (keeps Claude/Cursor responsive)	0
--verbose, -v	Print a notice each time auto-yes responds	off
--quiet, -q	Suppress all auto-yes output (no banner, no notices)	off
--pattern REGEX	Extra prompt pattern (repeatable)	—
--cli NAME	AI CLI profile to load (repeatable, or all)	—
--log FILE	Write debug log (raw output, cleaned text, detections)	—

---

Supported tools & pattern categories

AI CLI profiles

Profile	Binary	Tool	Key prompt patterns
claude	claude	Anthropic Claude Code	❯ 1. Yes, I trust this folder, ❯ Allow once, Do you want to proceed?
cursor	agent	Cursor Agent CLI	→ Run (once) (y), → Run (always) (a), ▶ [a] Trust this workspace
gemini	gemini	Google Gemini CLI	1. Allow once, 2. Allow for this session, 3. Always allow
codex	codex	OpenAI Codex CLI	1. Approve and run now, 1. Yes, allow Codex to work
copilot	gh copilot	GitHub Copilot CLI	1. Yes, proceed, Allow Copilot to run
aider	aider	Aider AI Coding	(Y)es/(N)o, Run shell command?, Add … to the chat?
openhands	openhands	OpenHands AI Agent	Do you want to execute this action?, Approve
windsurf	windsurf	Codeium Windsurf	Accept changes?, Run this command?
qwen	qwen	Alibaba Qwen Code	1. Yes, Approve execution?
amazonq	q	Amazon Q Developer	Do you approve this action?, Accept suggestion?
grok	grok	xAI Grok CLI	1. Yes
auggie	auggie	Augment Code CLI	[Y] Enable indexing
amp	amp	Sourcegraph Amp CLI	Approve

Generic patterns (opt-in via --cli generic)

Type	Examples
Bracket choices	[y/n], [Y/n], (y/N), [yes/no], ([y]/n), (y)
Styled choices	[Y]es / [N]o, bare y/n:
Question prompts	Continue?, Proceed?, Are you sure?, Would you like to…?
Safe action prompts	Allow?, Approve?, Accept?, Download?, Enable?, Create?
Agreement prompts	Do you agree?, Accept the license, Agree to the terms
SSH fingerprint	continue connecting (yes/no/[fingerprint])? → responds yes
Terraform	Only 'yes' will be accepted → responds yes
Full-word yes	Type 'yes' to continue, Enter 'yes' to proceed → responds yes
Press enter	Press Enter to continue → responds with empty string
Default value	(default: Y), [default=yes] → accepts default with empty
Excluded (dangerous)	Overwrite?, Delete?, Remove?, Upgrade?, Restart?, Reboot?

auto-yes patterns claude          # inspect Claude patterns
auto-yes patterns cursor gemini   # inspect multiple profiles
auto-yes patterns                 # list everything

Custom patterns

# persist a custom pattern for all future sessions
auto-yes add-pattern 'accept license\?'

# one-off extra pattern for a single run
auto-yes run --pattern 'custom_prompt\?' -- ./my-script.sh

---

Multi-agent & parallel agent scenarios

auto-yes is designed for multi-agent orchestration where multiple AI agents run concurrently and each may trigger prompts:

# 5 parallel Claude Code subagents — auto-yes handles all prompts
auto-yes run --auto -- orchestrator --agents 5 --task "migrate database schema"

# Overnight loop: run until complete, auto-approve every permission
while true; do
    auto-yes run --auto -- claude "continue where you left off"
    sleep 60
done

Tested with 5 simultaneous agents, 10-iteration overnight loops, interleaved Claude + Cursor prompts, and 1 000+ rapid sequential approvals.

---

Platform support

Platform	Method	Notes
Linux	pty + select	Full PTY, zero external deps
macOS	pty + select	Full PTY, zero external deps
Windows	pywinpty	Install auto-yes[windows]
Windows (fallback)	subprocess pipes	Works without pywinpty
Rocky Linux / RHEL	pty + select	Tested on Rocky 9
Ubuntu / Debian	pty + select	Tested on Ubuntu 22.04 / 24.04

---

Comparison with yes(1)

	yes | cmd	auto-yes run -- cmd
Prompt detection	None — floods stdin blindly	Smart regex, bottom-up scan
User can still type	No	Yes, fully interactive
PTY (colors / progress bars)	Broken	Preserved
False-positive protection	None	Typing suppress + cooldown
AI-specific prompts	No	13 AI tool profiles built-in
Custom patterns	No	Yes, persisted or one-off
Cross-platform	Unix only	Linux + macOS + Windows

---

Python API

from auto_yes.runner import Runner
from auto_yes.patterns import get_command

# wrap claude with all its patterns
runner = Runner(
    categories=["claude"],
    cooldown=0.1,
    verbose=True,
)
exit_code = runner.run_command(["claude", "fix the tests"])

# generic patterns for any command
runner = Runner(categories=["generic"], response="y")
exit_code = runner.run_command(["apt", "install", "nginx"])

# custom extra pattern on top of a profile
runner = Runner(
    categories=["cursor"],
    extra_patterns=[r"deploy to production\?"],
)
cmd = get_command("cursor")   # → ["agent"]
exit_code = runner.run_command(cmd + ["chat", "deploy the app"])

---

Configuration

Stored at ~/.config/auto-yes/config.json (Linux / macOS) or %APPDATA%\auto-yes\config.json (Windows):

{
  "custom_patterns": [],
  "response": "y",
  "cooldown": 0.5,
  "verbose": false
}

---

Adding a new AI CLI profile

Add one entry to REGISTRY in src/auto_yes/patterns.py:

_MY_TOOL = {
    "description": "My AI Tool CLI",
    "command": ["my-tool"],              # real binary (list of strings)
    "patterns": [
        (r"approve action\?", None),     # None → use default response ("y")
        (r"type yes to confirm", "yes"), # explicit response string
    ],
}
REGISTRY["my-tool"] = _MY_TOOL

No other file needs to change. Run auto-yes list to verify it appears.

---

Development

git clone https://github.com/host452b/auto-yes.git
cd auto-yes
pip install -e ".[dev]"

# run all 270 tests
pytest

# run only integration tests (real PTY scenarios)
pytest tests/scenarios/ -v

# run stress tests
pytest tests/test_stress.py -v

# lint + type check
ruff check src/ tests/
mypy src/auto_yes/

---

License

MIT © auto-yes contributors