All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
- Local Sandboxes: Isolated environments where AI coding agents can build software
runtm start- Start a sandbox with interactive mode/agent selectionruntm attach [id]- Attach to a sandbox (defaults to last active in terminal)runtm prompt "..."- Send prompts to the agent in autopilot moderuntm session list- List all sandbox sessionsruntm session stop <id>- Stop a sandbox (preserves workspace)runtm session destroy <id>- Destroy a sandbox and delete workspaceruntm session deploy- Deploy from sandbox to live URL
- Agent Orchestrator (
runtm-agents): New package for AI coding agent integration- Claude Code adapter with streaming JSON output parsing
- Autopilot mode: send prompts via CLI, agent executes autonomously
- Interactive mode: drop into sandbox shell, run agent manually
- Session continuation support (
--continueflag) - Real-time output streaming (tool use, text, errors)
- Sandbox Package (
runtm-sandbox): New package for sandbox management- Uses Anthropic's sandbox-runtime for fast startup (<100ms)
- OS-level isolation via bubblewrap (Linux) / seatbelt (macOS)
- Automatic dependency installation on first run
- Graceful fallback when ripgrep not installed
- Custom shell prompt showing sandbox ID
- Terminal-specific session tracking (multiple terminals, multiple sandboxes)
- Sandbox UX Improvements
- Welcome banner when entering sandbox
- Custom prompt:
[sandbox:abc123] ~/path $ - Exit message with next-step suggestions
- Environment variables for scripts:
RUNTM_SANDBOX,RUNTM_WORKSPACE
runtm listnow also available asruntm deployments list- Verbose logging now opt-in (
--verboseflag orRUNTM_DEBUG=1) - Keyring dependency now optional (falls back to file-based credential storage)
- Fixed
runtm startbypassing interactive menus - Fixed
runtm attachrequiring sandbox ID (now defaults to active session) - Fixed sandbox-runtime config format (
allowedDomainsvsallowDomains) - Fixed sandbox-runtime flag (
--settingsvs--config) - Fixed CLI crash when keyring package not installed
0.1.0 - 2025-01-01
- Initial open source release
- CLI:
runtmcommand-line tool for deploying AI-generated coderuntm init- Scaffold from templates (backend-service, static-site, web-app)runtm run- Run projects locally with auto-detection (uses Bun if available)runtm deploy- Deploy to live URLs with machine tiersruntm fix- Auto-fix common project issues (lockfiles)runtm validate- Validate projects before deploymentruntm status- Check deployment statusruntm logs- View build, deploy, and runtime logs with search/filteringruntm list- List all deploymentsruntm search- Search deployments by description/tagsruntm destroy- Destroy deploymentsruntm login/logout- Authentication managementruntm secrets set/get/list/unset- Manage environment variablesruntm domain add/status/remove- Custom domain managementruntm approve- Apply agent-proposed changesruntm admin create-token/revoke-token/list-tokens- Self-host token management
- API: FastAPI control plane for deployment management
- Worker: Build and deploy pipeline with Fly.io provider
- Templates:
backend-service- Python FastAPI backendstatic-site- Next.js static siteweb-app- Fullstack Next.js + FastAPI
- Features:
- Machine tiers (starter, standard, performance) with auto-stop
- Environment variable management with secret redaction
- Custom domain support with SSL certificates
- Optional SQLite database with persistence
- Optional authentication (web-app template)
- Agent workflow support via
runtm.requests.yaml - Lockfile validation and auto-fix
- Bearer token authentication for all API calls
- Rate limiting (10 deployments/hour per token)
- Artifact size limits (20 MB max)
- Build/deploy timeouts
- Secret redaction in logs