Changelog
We ship fast and often. Here's everything we've released.
Agents Dashboard: See Everything Your Agents Do
A dedicated page to manage all AI agents across your workspace. Browse registered and ad-hoc agents, monitor executions in real time, view detailed event streams, retry failed tasks, and track costs per run.
- Agents registry: browse all registered and ad-hoc agents across workspace apps
- Execution history with status, duration, cost, and token usage per run
- Live execution stream: watch agent tool calls, text output, and errors as they happen
- One-click retry for failed or partial executions
- Auto-refresh: execution list updates every 10s while agents are running
Skills Dashboard: Create, Import, and Share AI Skills
Manage workspace skills from a dedicated dashboard. Create custom skills with a markdown editor and live preview, import external skills, export and download skill files for use in any AI tool.
MCP Servers Hub: Health Checks, Tool Explorer, Quick Connect
A central hub for all MCP servers in your workspace. Live health monitoring with latency, a tool explorer to browse and test available tools, quick-connect with copyable endpoints, and API key management.
- Live health checks: connection status, latency, and tool count per server
- Tool explorer: browse and test MCP tools from any connected server
- Quick-connect: copyable MCP endpoint URLs with inline API key creation
- Integrations tab: see connected integrations as MCP tools
- Four-tab view: All, App MCPs, External Servers, Integrations
Task Delegation: Agents That Launch Agents
Agents can now delegate complex work to autonomous sub-agents running in sandboxes. Pass files in, get results back, set budgets and timeouts, and resume sessions across runs.
- run_agent_task tool: agents delegate work to sandboxed sub-agents
- Input and output files: pass data to task agents, collect results
- Budget caps, turn limits, and timeout controls per task
- Session resume: continue previous agent sessions with full context
- Cost tracking: input/output tokens and USD cost per execution
CLI: One Command to Connect 8 AI Agents
runwork setup now auto-detects and configures 8 AI agents: Claude Code, Claude Desktop, Cursor, Codex, Windsurf, Copilot (VSCode and CLI), and Gemini. One command to connect them all to your workspace skills and MCP servers.
Agent Sandbox: Your Agents Get Their Own Computer
App agents can now delegate complex tasks to Claude CLI running in isolated sandbox environments with full compute, file system, and workspace integration access.
Agent-Friendly CLI
The CLI now works better with coding agents like Claude Code, Cursor, and Codex. Clearer docs, machine-readable output, and smarter agent guidance built in.
Runwork CLI: code locally, ship globally
Install the CLI, run `runwork dev`, and build Runwork apps from your own editor with Claude Code, Cursor, Codex, or any coding agent you prefer. Everything stays in sync between local and web.
- Runwork CLI: install via npm, Homebrew, or curl and develop apps from your terminal
- `runwork dev`: develop apps locally with Claude Code, Cursor, Codex, or any editor and coding agent
- @runworkai/framework: typed SDK on npm with full autocomplete for your local coding environment
- Seamless local-to-web handover: start in your terminal, continue in the web dashboard, or vice versa
- Live preview sync: edit locally, see changes instantly in the cloud sandbox
- Welcome wizard with interactive agent guidance and keyboard shortcuts for fast onboarding
Smarter AI, Personal Integrations
App AI now supports images, tool use, and smarter caching. Plus, each team member can connect their own integration accounts.
- AI usage tracking revamped with transparent metering per workspace
The AI Ecosystem: your apps as Skills and MCP servers
Runwork apps can now expose Skills, serve as MCP servers, and connect to external MCP tools. Build AI-native workflows that talk to Claude, Cursor, and any MCP-compatible client.
- Skills: teach your apps domain-specific AI expertise with reusable skill definitions
- MCP Servers: expose any Runwork app as a tool server for Claude, Cursor, and other AI clients
- MCP Client: connect external MCP tool servers into your Runwork workflows
- Slash commands for quick actions inside the Work Assistant
Developer Power Tools
New developer-focused features for teams that need programmatic access and deeper control over their Runwork workspace.
- Improved code editor with syntax highlighting and autocomplete
- Fixed webhook delivery retries for failed endpoints
Workspace Channels
Real-time team communication built into your workspace. Channels combine chat, AI assistance, and app event feeds in one place.
- Workspace channels: team chat with threaded conversations
- AI in channels: ask the Work Assistant questions in any channel context
- App event feeds: see automation runs, errors, and deployments in dedicated channels
- Channel notifications with configurable alert levels
Build Faster
A faster path from idea to working app. Recipes let you start from proven templates, and the improved sandbox makes iteration instant.
- Recipes: browse and import from 8,000+ workflow templates with one click
- New sandbox environment with hot-reload preview
- Quick Edit mode for inline changes without opening the full editor
- Sharable Components: build once, reuse across apps in your workspace
Scale Up
Production infrastructure for growing teams. Usage-based billing, expanded integrations, and file handling for real-world workloads.
- Usage-based billing with transparent metering and spending controls
- File uploads and storage management for apps
Production Ready
Ship with confidence. App versioning, structured logs, built-in testing, and cross-app entity sharing make Runwork ready for production workloads.
- App versioning with rollback support
- Structured logging for automations and workflows
- Built-in testing framework for workflow and automation validation
- Cross-app entities: share data models between apps in the same workspace
- Improved error handling with actionable messages
Hello, Runwork
Analytics, admin controls, and a smoother onboarding experience to get teams productive faster.
- Built-in analytics dashboard for app usage and performance
- Admin roles with granular permission controls
- New onboarding flow with guided workspace setup
The Foundation
The initial release of the core platform. Workflows, team workspaces, the AI SDK, custom domains, and the command palette that ties it all together.
- Workflow engine with visual builder and code editor
- Team workspaces with member roles and shared resources
- AI SDK for building intelligent automations
- Custom domain support with automatic SSL
- Command palette for quick navigation and actions
Never miss an update
Subscribe to get notified when we ship new features.